Arxiv今日论文 | 2025-10-07

本篇博文主要内容为 2025-10-07 从Arxiv.org论文网站获取的最新论文列表，自动更新，按照NLP、CV、ML、AI、IR五个大方向区分，若需要邮件定时接收，请在评论区留下你的邮箱号。

说明：每日论文数据从Arxiv.org获取，每天早上12:00左右定时自动更新。

友情提示: 如何您需要邮箱接收每日论文数据，请在评论处留下你的邮箱。

【速读】：该论文旨在解决学术Presentation视频（即研究论文的可视化讲解视频）生成过程中存在的高劳动成本与多模态信息协调困难的问题。现有方法难以高效整合论文中的密集文本、图表、表格等多模态内容，并实现幻灯片、字幕、语音和虚拟人像等多通道的精准对齐。解决方案的关键在于提出首个面向学术视频生成的多智能体框架PaperTalker，其核心创新包括：基于有效树搜索的视觉选择机制（effective tree search visual choice）用于优化幻灯片布局，结合光标定位（cursor grounding）增强内容聚焦性，以及并行化逐页生成策略提升效率；同时构建了包含101篇论文及其配套视频、幻灯片和演讲者元数据的基准数据集，并设计Meta Similarity、PresentArena、PresentQuiz和IP Memory四项定制化评估指标，系统衡量视频对原始论文信息的忠实度与传播效果。实验表明，该方案显著优于现有基线，在自动化与实用性上迈出关键一步。

链接: https://arxiv.org/abs/2510.05096
作者: Zeyu Zhu,Kevin Qinghong Lin,Mike Zheng Shou
机构: Show Lab, National University of Singapore (新加坡国立大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Multiagent Systems (cs.MA); Multimedia (cs.MM)
备注: 20 pages, 8 figures

点击查看摘要

Abstract:Academic presentation videos have become an essential medium for research communication, yet producing them remains highly labor-intensive, often requiring hours of slide design, recording, and editing for a short 2 to 10 minutes video. Unlike natural video, presentation video generation involves distinctive challenges: inputs from research papers, dense multi-modal information (text, figures, tables), and the need to coordinate multiple aligned channels such as slides, subtitles, speech, and human talker. To address these challenges, we introduce PaperTalker, the first benchmark of 101 research papers paired with author-created presentation videos, slides, and speaker metadata. We further design four tailored evaluation metrics–Meta Similarity, PresentArena, PresentQuiz, and IP Memory–to measure how videos convey the paper’s information to the audience. Building on this foundation, we propose PaperTalker, the first multi-agent framework for academic presentation video generation. It integrates slide generation with effective layout refinement by a novel effective tree search visual choice, cursor grounding, subtitling, speech synthesis, and talking-head rendering, while parallelizing slide-wise generation for efficiency. Experiments on Paper2Video demonstrate that the presentation videos produced by our approach are more faithful and informative than existing baselines, establishing a practical step toward automated and ready-to-use academic video generation. Our dataset, agent, and code are available at this https URL.
zh

[NLP-1] From Noisy Traces to Stable Gradients: Bias-Variance Optimized Preference Optimization for Aligning Large Reasoning Models

【速读】：该论文旨在解决大推理模型（Large Reasoning Models, LRMs）在偏好对齐（preference alignment）过程中因推理轨迹（reasoning traces）采样带来的梯度方差过大问题。由于最优的偏好对齐目标需对所有可能的推理轨迹进行边际化，而这一计算在实践中不可行，现有方法通常仅优化单条采样轨迹，导致梯度估计方差显著增加，进而影响训练稳定性与模型性能。解决方案的关键在于将偏好优化建模为偏差-方差权衡（bias-variance trade-off）问题，并提出偏差-方差优化偏好优化（Bias-Variance Optimized Preference Optimization, BVPO），通过混合两种梯度估计器——高方差的基于轨迹的估计器与低方差的空轨迹估计器（禁用推理轨迹生成得到）——来有效降低梯度方差。理论分析表明，BVPO在任意非平凡混合比例下均可严格减少由轨迹采样引起的方差，并提供最小均方误差意义下的最优混合权重；实验验证其在AlpacaEval~2和Arena-Hard等基准上相比最佳基线提升最高达7.8和6.8点，同时显著提升基础模型在数学推理任务上的表现（最高提升4.0点）。

链接: https://arxiv.org/abs/2510.05095
作者: Mingkang Zhu,Xi Chen,Bei Yu,Hengshuang Zhao,Jiaya Jia
机构: The Chinese University of Hong Kong (香港中文大学); The University of Hong Kong (香港大学); The Hong Kong University of Science and Technology (香港科技大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:

点击查看摘要

Abstract:Large reasoning models (LRMs) generate intermediate reasoning traces before producing final answers, yielding strong gains on multi-step and mathematical tasks. Yet aligning LRMs with human preferences, a crucial prerequisite for model deployment, remains underexplored. The statistically correct objective for preference alignment requires marginalizing over reasoning traces, but this computation is intractable in practice. A common workaround optimizes a single sampled trajectory, which introduces substantial gradient variance from stochastic trace sampling. To address this challenge, we frame preference optimization for LRMs through the lens of the bias–variance trade-off and propose Bias–Variance Optimized Preference Optimization (BVPO), a simple, drop-in method that mixes two gradient estimators: a high-variance trace-based estimator and a low-variance empty-trace estimator obtained by disabling reasoning trace generation. Our theory shows that BVPO strictly reduces trace-induced variance for any nontrivial mixture, provides a closed-form choice of the mixing weight that minimizes mean-squared error relative to the true marginal gradient, and under standard smoothness and step-size conditions, tightens classical convergence bounds for stochastic gradient descent. Empirically, BVPO improves alignment over the best baseline by up to 7.8 points on AlpacaEval~2 and 6.8 points on Arena-Hard. Despite being trained only on general conversational data, BVPO also boosts reasoning performance for base models by up to 4.0 points on the average of six math reasoning benchmarks. These results identify variance from trace sampling as a key bottleneck and demonstrate that directly optimizing the bias–variance trade-off yields more stable training and stronger overall performance.
zh

[NLP-2] Learning to Interpret Weight Differences in Language Models

【速读】：该论文旨在解决预训练语言模型在微调（finetuning）过程中产生的权重变化（weight diffs）难以解释的问题。现有方法通常依赖于对微调数据集的分析来推测模型变化，但这些数据集往往不可公开或规模过大，无法直接使用。解决方案的关键在于提出一种名为“权重差异解释微调”（Diff Interpretation Tuning, DIT）的方法：通过合成且标注的权重差异数据训练一个DIT适配器（adapter），该适配器可应用于兼容的微调模型，使其以自然语言准确描述自身因微调所引发的修改。实验表明，DIT能够在两个概念验证场景中（如报告隐藏行为、总结微调知识）实现对权重变化的可解释性输出。

链接: https://arxiv.org/abs/2510.05092
作者: Avichal Goel,Yoon Kim,Nir Shavit,Tony T. Wang
机构: Massachusetts Institute of Technology (麻省理工学院)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: The weight diffs and DIT adapters trained in the paper can be found at this https URL

点击查看摘要

Abstract:Finetuning (pretrained) language models is a standard approach for updating their internal parametric knowledge and specializing them to new tasks and domains. However, the corresponding model weight changes (“weight diffs”) are not generally interpretable. While inspecting the finetuning dataset can give a sense of how the model might have changed, these datasets are often not publicly available or are too large to work with directly. Towards the goal of comprehensively understanding weight diffs in natural language, we introduce Diff Interpretation Tuning (DIT), a method that trains models to describe their own finetuning-induced modifications. Our approach uses synthetic, labeled weight diffs to train a DIT adapter, which can be applied to a compatible finetuned model to make it describe how it has changed. We demonstrate in two proof-of-concept settings (reporting hidden behaviors and summarizing finetuned knowledge) that our method enables models to describe their finetuning-induced modifications using accurate natural language descriptions.
zh

[NLP-3] Finish First Perfect Later: Test-Time Token-Level Cross-Validation for Diffusion Large Language Models

链接: https://arxiv.org/abs/2510.05090
作者: Runchu Tian,Junxia Cui,Xueqiang Xu,Feng Yao,Jingbo Shang
机构: University of Illinois Urbana-Champaign (伊利诺伊大学厄巴纳-香槟分校); University of California San Diego (加州大学圣地亚哥分校)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 17 pages, 8 figures. Work in progress

点击查看摘要

[NLP-4] achLM: Post-Training LLM s for Education Using Authentic Learning Data

【速读】：该论文旨在解决生成式 AI 在教育领域应用中因大型语言模型（Large Language Models, LLMs）教学能力受限的问题，尤其是缺乏反映真实学生学习过程的高质量训练数据，以及基于规则的提示工程在编码复杂教学策略方面的固有局限。解决方案的关键在于提出 TeachLM——一个通过参数高效微调（Parameter-Efficient Fine-Tuning）优化的教学专用大模型，其训练数据来源于 Polygence 平台维护的 10 万小时一对一、纵向的学生-导师互动记录，并经过严格匿名化处理以保障隐私。该方法构建了一个高保真度的合成学生模型，从而支持生成逼真的师生对话，进一步推动了一种新颖的多轮对话评估协议，实现对 LLM 教学对话能力的快速、可扩展且可复现的测评。实验表明，基于真实学习数据的微调显著提升了对话连贯性与教学效果：学生发言时长翻倍、提问方式优化、对话轮次增加 50%，并增强个性化教学水平。

链接: https://arxiv.org/abs/2510.05087
作者: Janos Perczel,Jin Chow,Dorottya Demszky
机构: Polygence; Stanford University (斯坦福大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 28 pages, 9 figures

点击查看摘要

Abstract:The promise of generative AI to revolutionize education is constrained by the pedagogical limits of large language models (LLMs). A major issue is the lack of access to high-quality training data that reflect the learning of actual students. Prompt engineering has emerged as a stopgap, but the ability of prompts to encode complex pedagogical strategies in rule-based natural language is inherently limited. To address this gap we introduce TeachLM - an LLM optimized for teaching through parameter-efficient fine-tuning of state-of-the-art models. TeachLM is trained on a dataset comprised of 100,000 hours of one-on-one, longitudinal student-tutor interactions maintained by Polygence, which underwent a rigorous anonymization process to protect privacy. We use parameter-efficient fine-tuning to develop an authentic student model that enables the generation of high-fidelity synthetic student-tutor dialogues. Building on this capability, we propose a novel multi-turn evaluation protocol that leverages synthetic dialogue generation to provide fast, scalable, and reproducible assessments of the dialogical capabilities of LLMs. Our evaluations demonstrate that fine-tuning on authentic learning data significantly improves conversational and pedagogical performance - doubling student talk time, improving questioning style, increasing dialogue turns by 50%, and greater personalization of instruction.
zh

[NLP-5] Slm-mux: Orchestrating small language models for reasoning

链接: https://arxiv.org/abs/2510.05077
作者: Chenyu Wang,Zishen Wan,Hao Kang,Emma Chen,Zhiqiang Xie,Tushar Krishna,Vijay Janapa Reddi,Yilun Du
机构: Harvard University (哈佛大学); Georgia Institute of Technology (佐治亚理工学院); Stanford University (斯坦福大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[NLP-6] SwiReasoning : Switch-Thinking in Latent and Explicit for Pareto-Superior Reasoning LLM s

【速读】：该论文旨在解决大语言模型（Large Language Models, LLMs）在无需训练的条件下进行潜空间推理（latent reasoning）时面临的两个核心问题：一是纯潜空间推理因维持多个隐式路径而导致概率质量分散、引入噪声并阻碍收敛，从而降低准确性；二是即使不生成显式文本，仍存在过度思考（overthinking）现象，浪费计算资源并降低令牌效率（token efficiency）。解决方案的关键在于提出一种无需训练的框架 SwiReasoning，其核心创新包括：1）基于下一标记分布熵趋势估算的块级置信度，动态切换显式与潜空间推理，以平衡探索与利用并促进及时收敛；2）通过限制最大思维块切换次数，有效抑制过度思考，在不同难度的问题上提升令牌效率。实验证明，该方法在数学与STEM基准测试中平均准确率提升1.5%-2.8%，在有限预算下令牌效率提升56%-79%。

链接: https://arxiv.org/abs/2510.05069
作者: Dachuan Shi,Abedelkadir Asi,Keying Li,Xiangchi Yuan,Leyan Pan,Wenke Lee,Wen Xiao
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Code: this https URL , Website: this https URL

点击查看摘要

Abstract:Recent work shows that, beyond discrete reasoning through explicit chain-of-thought steps, which are limited by the boundaries of natural languages, large language models (LLMs) can also reason continuously in latent space, allowing richer information per step and thereby improving token efficiency. Despite this promise, latent reasoning still faces two challenges, especially in training-free settings: 1) purely latent reasoning broadens the search distribution by maintaining multiple implicit paths, which diffuses probability mass, introduces noise, and impedes convergence to a single high-confidence solution, thereby hurting accuracy; and 2) overthinking persists even without explicit text, wasting tokens and degrading efficiency. To address these issues, we introduce SwiReasoning, a training-free framework for LLM reasoning which features two key innovations: 1) SwiReasoning dynamically switches between explicit and latent reasoning, guided by block-wise confidence estimated from entropy trends in next-token distributions, to balance exploration and exploitation and promote timely convergence. 2) By limiting the maximum number of thinking-block switches, SwiReasoning curbs overthinking and improves token efficiency across varying problem difficulties. On widely used mathematics and STEM benchmarks, SwiReasoning consistently improves average accuracy by 1.5%-2.8% across reasoning LLMs of different model families and scales. Furthermore, under constrained budgets, SwiReasoning improves average token efficiency by 56%-79%, with larger gains as budgets tighten.
zh

[NLP-7] Proactive defense against LLM Jailbreak

【速读】：该论文旨在解决大语言模型（Large Language Models, LLMs）在面对持续演进的对抗性攻击（尤其是多轮越狱攻击，multi-turn jailbreaks）时安全性不足的问题。这类攻击通过迭代搜索成功查询来绕过安全机制，而现有以静态、被动响应为主的防御方法难以有效应对。其解决方案的关键在于提出一种名为ProAct的主动防御框架：该框架通过有意识地生成看似成功的“虚假响应”（spurious responses），误导攻击者的内部优化循环，使其误认为已达成越狱目标从而提前终止攻击，进而实现对越狱过程的“反越狱”。实验表明，该方法可将攻击成功率显著降低至最高92%，并与现有防御机制结合后使最新攻击策略的成功率降至0%。

链接: https://arxiv.org/abs/2510.05052
作者: Weiliang Zhao,Jinjun Peng,Daniel Ben-Levi,Zhou Yu,Junfeng Yang
机构: Columbia University (哥伦比亚大学)
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL)
备注:

点击查看摘要

Abstract:The proliferation of powerful large language models (LLMs) has necessitated robust safety alignment, yet these models remain vulnerable to evolving adversarial attacks, including multi-turn jailbreaks that iteratively search for successful queries. Current defenses, primarily reactive and static, often fail to counter these search-based attacks. In this paper, we introduce ProAct, a novel proactive defense framework designed to disrupt and mislead autonomous jailbreaking processes. Our core idea is to intentionally provide adversaries with “spurious responses” that appear to be results of successful jailbreak attacks but contain no actual harmful content. These misleading responses provide false signals to the attacker’s internal optimization loop, causing the adversarial search to terminate prematurely and effectively jailbreaking the jailbreak. By conducting extensive experiments across state-of-the-art LLMs, jailbreaking frameworks, and safety benchmarks, our method consistently and significantly reduces attack success rates by up to 92%. When combined with other defense frameworks, it further reduces the success rate of the latest attack strategies to 0%. ProAct represents an orthogonal defense strategy that can serve as an additional guardrail to enhance LLM safety against the most effective jailbreaking attacks.
zh

[NLP-8] COLE: a Comprehensive Benchmark for French Language Understanding Evaluation ACL

链接: https://arxiv.org/abs/2510.05046
作者: David Beauchemin,Yan Tremblay,Mohamed Amine Youssef,Richard Khoury
机构: 未知
类目: Computation and Language (cs.CL)
备注: Submitted to ACL Rolling Review of October

点击查看摘要

[NLP-9] Guided Query Refinement: Multimodal Hybrid Retrieval with Test-Time Optimization

链接: https://arxiv.org/abs/2510.05038
作者: Omri Uzan,Asaf Yehudai,Roi pony,Eyal Shnarch,Ariel Gera
机构: Stanford University (斯坦福大学); IBM Research; The Hebrew University of Jerusalem (希伯来大学)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-10] A Set of Quebec-French Corpus of Regional Expressions and Terms ACL

【速读】：该论文旨在解决当前自然语言处理（Natural Language Processing, NLP）中对区域性语言特征（如方言和习语）理解能力评估不足的问题。传统上，习语理解和方言理解被视为独立的基准任务，但缺乏将二者结合以更精细地衡量模型在特定方言下语言能力的方法。解决方案的关键在于提出两个新的基准数据集——QFrCoRE（含4,633个习语短语实例）和QFrCoRT（含171个魁北克法语地区性习语词汇实例），并利用区域习语作为方言理解的测试指标，从而为评估大语言模型（Large Language Models, LLMs）在特定方言中的表现提供可靠工具。该方法具有可复制性，可推广至其他方言的语言建模与评测场景。

链接: https://arxiv.org/abs/2510.05026
作者: David Beauchemin,Yan Tremblay,Mohamed Amine Youssef,Richard Khoury
机构: 未知
类目: Computation and Language (cs.CL)
备注: Submitted to ACL Rolling Review of October

点击查看摘要

Abstract:The tasks of idiom understanding and dialect understanding are both well-established benchmarks in natural language processing. In this paper, we propose combining them, and using regional idioms as a test of dialect understanding. Towards this end, we propose two new benchmark datasets for the Quebec dialect of French: QFrCoRE, which contains 4,633 instances of idiomatic phrases, and QFrCoRT, which comprises 171 regional instances of idiomatic words. We explain how to construct these corpora, so that our methodology can be replicated for other dialects. Our experiments with 94 LLM demonstrate that our regional idiom benchmarks are a reliable tool for measuring a model’s proficiency in a specific dialect.
zh

[NLP-11] Imperceptible Jailbreaking against Large Language Models

链接: https://arxiv.org/abs/2510.05025
作者: Kuofeng Gao,Yiming Li,Chao Du,Xin Wang,Xingjun Ma,Shu-Tao Xia,Tianyu Pang
机构: Tsinghua University (清华大学); Sea AI Lab, Singapore (新加坡海AI实验室); Nanyang Technological University (南洋理工大学); Fudan University (复旦大学); Peng Cheng Laboratory (鹏城实验室)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:

点击查看摘要

[NLP-12] Resource-Efficient Fine-Tuning of LLaMA-3.2-3B for Medical Chain-of-Thought Reasoning

链接: https://arxiv.org/abs/2510.05003
作者: Imran Mansha
机构: The Islamia University of Bahawalpur (巴基斯坦巴哈瓦尔布尔伊斯兰大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 6 pages, 2 figures. Submitted to arXiv for open access

点击查看摘要

[NLP-13] Reinforce-Ada: An Adaptive Sampling Framework for Reinforce-Style LLM Training

链接: https://arxiv.org/abs/2510.04996
作者: Wei Xiong,Chenlu Ye,Baohao Liao,Hanze Dong,Xinxing Xu,Christof Monz,Jiang Bian,Nan Jiang,Tong Zhang
机构: University of Illinois Urbana-Champaign (伊利诺伊大学厄巴纳-香槟分校); Microsoft Research (微软研究院); University of Amsterdam (阿姆斯特丹大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (stat.ML)
备注: 16 pages, 6 figures

点击查看摘要

[NLP-14] AWARE Beyond Sentence Boundaries: A Contextual Transformer Framework for Identifying Cultural Capital in STEM Narratives

链接: https://arxiv.org/abs/2510.04983
作者: Khalid Mehtab Khan,Anagha Kulkarni
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Machine Learning (cs.LG)
备注:

点击查看摘要

[NLP-15] LLM -Hanabi: Evaluating Multi-Agent Gameplays with Theory-of-Mind and Rationale Inference in Imperfect Information Collaboration Game EMNLP2025

链接: https://arxiv.org/abs/2510.04980
作者: Fangzhou Liang,Tianshi Zheng,Chunkit Chan,Yauwai Yim,Yangqiu Song
机构: HKUST(香港科技大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: EMNLP 2025 Wordplay

点击查看摘要

[NLP-16] Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy (short paper) ACL2025

链接: https://arxiv.org/abs/2510.04950
作者: Om Dobariya,Akhil Kumar
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE); Methodology (stat.ME)
备注: 5 pages, 3 tables; includes Limitations and Ethical Considerations sections; short paper under submission to Findings of ACL 2025

点击查看摘要

[NLP-17] A First Context-Free Grammar Applied to Nawatl Corpora Augmentation

链接: https://arxiv.org/abs/2510.04945
作者: Juan-José Guzmán-Landa,Juan-Manuel Torres-Moreno,Miguel Figueroa-Saavedra,Ligia Quintana-Torres,Martha-Lorena Avendaño-Garrido,Graham Ranger
机构: Avignon Université (阿维尼翁大学); Universidad Veracruzana (韦拉克鲁斯大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 11 pages, 7 tables, 1 figure

点击查看摘要

[NLP-18] On Structured State-Space Duality

【速读】：该论文旨在解决结构化状态空间模型（Structured State-Space Model, SSM）与掩码自注意力机制之间的等价性问题，特别是如何在保持计算效率的同时扩展模型的表达能力。其核心贡献在于形式化并推广了结构化状态空间对偶性（Structured State-Space Duality, SSD），关键解决方案是将原始仅限于标量-单位矩阵状态矩阵的SSM扩展至一般对角线SSM，并证明这类模型在支持更丰富动态的同时仍能匹配标量情形下的训练复杂度下界；同时，论文给出了SSM等价于1-半分离掩码注意力的充要条件，并指出该对偶性无法推广到标准softmax注意力，因其存在秩爆炸问题。这一成果深化了递归SSM与Transformer架构之间的理论联系，并拓宽了高效且具有表现力的序列建模设计空间。

链接: https://arxiv.org/abs/2510.04944
作者: Jerry Yao-Chieh Hu,Xiwen Zhang,Weimin Wu,Han Liu
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (stat.ML)
备注:

点击查看摘要

Abstract:Structured State-Space Duality (SSD) [Dao Gu, ICML 2024] is an equivalence between a simple Structured State-Space Model (SSM) and a masked attention mechanism. In particular, a state-space model with a scalar-times-identity state matrix is equivalent to a masked self-attention with a 1 -semiseparable causal mask. Consequently, the same sequence transformation (model) has two algorithmic realizations: as a linear-time O(T) recurrence or as a quadratic-time O(T^2) attention. In this note, we formalize and generalize this duality: (i) we extend SSD from the scalar-identity case to general diagonal SSMs (diagonal state matrices); (ii) we show that these diagonal SSMs match the scalar case’s training complexity lower bounds while supporting richer dynamics; (iii) we establish a necessary and sufficient condition under which an SSM is equivalent to 1 -semiseparable masked attention; and (iv) we show that such duality fails to extend to standard softmax attention due to rank explosion. Together, these results tighten bridge between recurrent SSMs and Transformers, and widen the design space for expressive yet efficient sequence models.
zh

[NLP-19] ONNX-Net: Towards Universal Representations and Instant Performance Prediction for Neural Architectures

链接: https://arxiv.org/abs/2510.04938
作者: Shiwen Qin,Alexander Auras,Shay B. Cohen,Elliot J. Crowley,Michael Moeller,Linus Ericsson,Jovita Lukasik
机构: University of Edinburgh (爱丁堡大学); University of Siegen (锡根大学); University of Glasgow (格拉斯哥大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Our code is available at: this https URL

点击查看摘要

[NLP-20] MARS: Optimizing Dual-System Deep Research via Multi-Agent Reinforcement Learning

链接: https://arxiv.org/abs/2510.04935
作者: Guoxin Chen,Zile Qiao,Wenqing Wang,Donglei Yu,Xuanzhong Chen,Hao Sun,Minpeng Liao,Kai Fan,Yong Jiang,Penguin Xie,Wayne Xin Zhao,Ruihua Song,Fei Huang
机构: Gaoling School of Artificial Intelligence, Renmin University of China (中国人民大学高瓴人工智能学院); Tongyi Lab, Alibaba Group (阿里巴巴集团通义实验室)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Ongoing Work

点击查看摘要

[NLP-21] he Geometry of Truth: Layer-wise Semantic Dynamics for Hallucination Detection in Large Language Models

链接: https://arxiv.org/abs/2510.04933
作者: Amir Hameed Mir
机构: Sirraya Labs
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Theory (cs.IT); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
备注: Comments: 14 pages, 14 figures, 5 tables. Code available at: this https URL

点击查看摘要

[NLP-22] Do LLM s Align with My Task? Evaluating Text-to-SQL via Dataset Alignment

链接: https://arxiv.org/abs/2510.04919
作者: Davood Rafiei,Morgan Lindsay Heisler,Weiwei Zhang,Mohammadreza Pourreza,Yong Zhang
机构: University of Alberta (阿尔伯塔大学); Huawei Tech. Canada (华为技术加拿大)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Databases (cs.DB)
备注:

点击查看摘要

[NLP-23] Retrieval-Augmented Code Generation: A Survey with Focus on Repository-Level Approaches

【速读】：该论文旨在解决**仓库级代码生成（Repository-Level Code Generation, RLCG）这一挑战性问题，即在真实软件开发场景中，模型需跨越多个文件或模块进行推理，确保长距离依赖关系的捕捉、全局语义一致性以及跨文件代码的连贯性。其解决方案的关键在于引入检索增强生成（Retrieval-Augmented Generation, RAG）**范式，通过整合外部检索机制与大语言模型（Large Language Models, LLMs），显著提升模型对上下文的理解能力与可扩展性，从而实现更高质量的仓库级代码生成。

链接: https://arxiv.org/abs/2510.04905
作者: Yicheng Tao,Yao Qin,Yepang Liu
机构: Carnegie Mellon University (卡内基梅隆大学); Chinese University of Hong Kong (香港中文大学); Southern University of Science and Technology (南方科技大学)
类目: oftware Engineering (cs.SE); Computation and Language (cs.CL)
备注:

点击查看摘要

Abstract:Recent advancements in large language models (LLMs) have substantially improved automated code generation. While function-level and file-level generation have achieved promising results, real-world software development typically requires reasoning across entire repositories. This gives rise to the challenging task of Repository-Level Code Generation (RLCG), where models must capture long-range dependencies, ensure global semantic consistency, and generate coherent code spanning multiple files or modules. To address these challenges, Retrieval-Augmented Generation (RAG) has emerged as a powerful paradigm that integrates external retrieval mechanisms with LLMs, enhancing context-awareness and scalability. In this survey, we provide a comprehensive review of research on Retrieval-Augmented Code Generation (RACG), with an emphasis on repository-level approaches. We categorize existing work along several dimensions, including generation strategies, retrieval modalities, model architectures, training paradigms, and evaluation protocols. Furthermore, we summarize widely used datasets and benchmarks, analyze current limitations, and outline key challenges and opportunities for future research. Our goal is to establish a unified analytical framework for understanding this rapidly evolving field and to inspire continued progress in AI-powered software engineering.
zh

[NLP-24] SocialHarmBench: Revealing LLM Vulnerabilities to Socially Harmful Requests

链接: https://arxiv.org/abs/2510.04891
作者: Punya Syon Pandey,Hai Son Le,Devansh Bhardwaj,Rada Mihalcea,Zhijing Jin
机构: University of Toronto (多伦多大学); Vector Institute (向量研究所); Toronto Metropolitan University (多伦多都会大学); IIT, Roorkee (印度理工学院，鲁尔基分校); University of Michigan (密歇根大学); MPI for Intelligent Systems, Tübingen, Germany (马克斯·普朗克智能系统研究所，德国图宾根)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[NLP-25] Detecting Distillation Data from Reasoning Models

链接: https://arxiv.org/abs/2510.04850
作者: Hengxiang Zhang,Hyeong Kyu Choi,Yixuan Li,Hongxin Wei
机构: Southern University of Science and Technology (南方科技大学); University of Wisconsin–Madison (威斯康星大学麦迪逊分校)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[NLP-26] When Models Lie We Learn: Multilingual Span-Level Hallucination Detection with PsiloQA

链接: https://arxiv.org/abs/2510.04849
作者: Elisei Rykov,Kseniia Petrushina,Maksim Savkin,Valerii Olisov,Artem Vazhentsev,Kseniia Titova,Alexander Panchenko,Vasily Konovalov,Julia Belikova
机构: Skoltech(斯科尔科沃科学技术研究所); AIRI(人工智能研究学院); MWS AI; Sber AI Lab(斯贝AI实验室); Moscow Institute of Physics and Technology(莫斯科物理技术学院)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-27] Instability in Downstream Task Performance During LLM Pretraining EMNLP2025

链接: https://arxiv.org/abs/2510.04848
作者: Yuto Nishida,Masaru Isonuma,Yusuke Oda
机构: Nara Institute of Science and Technology (奈良科学技术大学院大学); Tohoku University (东北大学); Research and Development Center for Large Language Models, National Institute of Informatics (日本信息研究所大语言模型研发中心)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2025 Findings

点击查看摘要

[NLP-28] How I Built ASR for Endangered Languages with a Spoken Dictionary

链接: https://arxiv.org/abs/2510.04832
作者: Christopher Bartley,Anton Ragni
机构: 未知
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-29] Visual Representations inside the Language Model

链接: https://arxiv.org/abs/2510.04819
作者: Benlin Liu,Amita Kamath,Madeleine Grunde-McLaughlin,Winson Han,Ranjay Krishna
机构: University of Washington (华盛顿大学); University of California Los Angeles (加州大学洛杉矶分校); Allen Institute for AI (艾伦人工智能研究所)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: Accepted to COLM 2025

点击查看摘要

[NLP-30] Hybrid Architectures for Language Models: Systematic Analysis and Design Insights

链接: https://arxiv.org/abs/2510.04800
作者: Sangmin Bae,Bilge Acun,Haroun Habeeb,Seungyeon Kim,Chien-Yu Lin,Liang Luo,Junjie Wang,Carole-Jean Wu
机构: KAIST (韩国科学技术院); Meta (Meta)
类目: Computation and Language (cs.CL)
备注: 17 pages, 4 figures, 6 tables; detailed results will be included in the Appendix later

点击查看摘要

[NLP-31] Are BabyLMs Deaf to Gricean Maxims? A Prag matic Evaluation of Sample-efficient Language Models

链接: https://arxiv.org/abs/2510.04764
作者: Raha Askari,Sina Zarrieß,Özge Alacam,Judith Sieker
机构: University of Turin (都灵大学); Bielefeld University (比勒费尔德大学)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-32] ModernBERT ColBERT: Enhancing biomedical RAG through an advanced re-ranking retriever

链接: https://arxiv.org/abs/2510.04757
作者: Eduardo Martínez Rivera,Filippo Menolascina
机构: University of Edinburgh (爱丁堡大学)
类目: Computation and Language (cs.CL); Quantitative Methods (q-bio.QM)
备注:

点击查看摘要

[NLP-33] A Low-Resource Speech-Driven NLP Pipeline for Sinhala Dyslexia Assistance

【速读】：该论文旨在解决成人阅读障碍（dyslexia）在非英语语境下研究不足与支持匮乏的问题，尤其针对斯里兰卡的僧伽罗语（Sinhala）这一低资源语言缺乏可访问性工具的现状。其解决方案的关键在于构建一个端到端的多模态辅助系统：利用Whisper实现语音转文字，基于SinBERT模型识别僧伽罗语中的常见阅读障碍错误，结合mT5与Mistral模型生成修正文本，并通过gTTS将修正后的文本重新转换为语音，形成闭环反馈。该系统在有限数据条件下实现了0.66的转录准确率和0.7的纠错准确率，验证了面向低资源语言的包容性自然语言处理（Natural Language Processing, NLP）技术的可行性与有效性。

链接: https://arxiv.org/abs/2510.04750
作者: Peshala Perera,Deshan Sumanathilaka
机构: 未知
类目: Computation and Language (cs.CL); Software Engineering (cs.SE)
备注: 11 pages, 4 figures, 3 tables

点击查看摘要

Abstract:Dyslexia in adults remains an under-researched and under-served area, particularly in non-English-speaking contexts, despite its significant impact on personal and professional lives. This work addresses that gap by focusing on Sinhala, a low-resource language with limited tools for linguistic accessibility. We present an assistive system explicitly designed for Sinhala-speaking adults with dyslexia. The system integrates Whisper for speech-to-text conversion, SinBERT, an open-sourced fine-tuned BERT model trained for Sinhala to identify common dyslexic errors, and a combined mT5 and Mistral-based model to generate corrected text. Finally, the output is converted back to speech using gTTS, creating a complete multimodal feedback loop. Despite the challenges posed by limited Sinhala-language datasets, the system achieves 0.66 transcription accuracy and 0.7 correction accuracy with 0.65 overall system accuracy. These results demonstrate both the feasibility and effectiveness of the approach. Ultimately, this work highlights the importance of inclusive Natural Language Processing (NLP) technologies in underrepresented languages and showcases a practical
zh

[NLP-34] Speak Edit Repeat: High-Fidelity Voice Editing and Zero-Shot TTS with Cross-Attentive Mamba

链接: https://arxiv.org/abs/2510.04738
作者: Baher Mohammad,Magauiya Zhussip,Stamatios Lefkimmiatis
机构: MTS AI; ITMO University
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS)
备注:

点击查看摘要

[NLP-35] BrokenMath: A Benchmark for Sycophancy in Theorem Proving with LLM s

链接: https://arxiv.org/abs/2510.04721
作者: Ivo Petrov,Jasper Dekoninck,Martin Vechev
机构: INSAIT; Sofia University “St. Kliment Ohridski”; ETH Zurich
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:

点击查看摘要

[NLP-36] JSON Whisperer: Efficient JSON Editing with LLM s

链接: https://arxiv.org/abs/2510.04717
作者: Sarel Duanis,Asnat Greenstein-Messica,Eliya Habba
机构: Lightricks; The Hebrew University of Jerusalem
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-37] Multilingual Routing in Mixture-of-Experts

链接: https://arxiv.org/abs/2510.04694
作者: Lucas Bandarkar,Chenyuan Yang,Mohsen Fayyaz,Junlin Hu,Nanyun Peng
机构: University of California, Los Angeles (加州大学洛杉矶分校); Fudan University (复旦大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[NLP-38] ok: Transfer Token-level Knowledge via Contrastive Excess to Transplant LoRA

链接: https://arxiv.org/abs/2510.04682
作者: Chanjoo Jung,Jaehyung Kim
机构: Yonsei University (延世大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[NLP-39] Multi-Agent Tool-Integrated Policy Optimization

链接: https://arxiv.org/abs/2510.04678
作者: Zhanfeng Mo,Xingxuan Li,Yuntao Chen,Lidong Bing
机构: 未知
类目: Computation and Language (cs.CL)
备注: Work in progress

点击查看摘要

[NLP-40] FocusMed: A Large Language Model-based Framework for Enhancing Medical Question Summarization with Focus Identification

【速读】：该论文旨在解决在线医疗平台中消费者健康问题（Consumer Health Questions, CHQs）因冗余信息和非专业术语导致的诊断效率低下问题，具体聚焦于医学问题摘要（Medical Question Summary, MQS）任务中的两大挑战：问题焦点识别不准与模型幻觉（hallucination）。解决方案的关键在于提出一种基于核心焦点引导的优化框架：首先设计提示模板以驱动大语言模型（Large Language Models, LLMs）忠实提取CHQ的核心焦点；其次结合原始CHQ-FAQ配对构建微调数据集，增强模型对问题焦点的识别能力；最后引入多维质量评估与选择机制，从多个维度系统提升摘要质量。实验表明，该框架在两个主流MQS数据集上均达到最优性能，显著提升了焦点识别准确性并有效缓解了幻觉现象。

链接: https://arxiv.org/abs/2510.04671
作者: Chao Liu,Ling Luo,Tengxiao Lv,Huan Zhuang,Lejing Yu,Jian Wang,Hongfei Lin
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted as a regular paper at BIBM2025

点击查看摘要

Abstract:With the rapid development of online medical platforms, consumer health questions (CHQs) are inefficient in diagnosis due to redundant information and frequent non-professional terms. The medical question summary (MQS) task aims to transform CHQs into streamlined doctors’ frequently asked questions (FAQs), but existing methods still face challenges such as poor identification of question focus and model hallucination. This paper explores the potential of large language models (LLMs) in the MQS task and finds that direct fine-tuning is prone to focus identification bias and generates unfaithful content. To this end, we propose an optimization framework based on core focus guidance. First, a prompt template is designed to drive the LLMs to extract the core focus from the CHQs that is faithful to the original text. Then, a fine-tuning dataset is constructed in combination with the original CHQ-FAQ pairs to improve the ability to identify the focus of the question. Finally, a multi-dimensional quality evaluation and selection mechanism is proposed to comprehensively improve the quality of the summary from multiple dimensions. We conduct comprehensive experiments on two widely-adopted MQS datasets using three established evaluation metrics. The proposed framework achieves state-of-the-art performance across all measures, demonstrating a significant boost in the model’s ability to identify critical focus of questions and a notable mitigation of hallucinations. The source codes are freely available at this https URL.
zh

[NLP-41] FT-MDT: Extracting Decision Trees from Medical Texts via a Novel Low-rank Adaptation Method EMNLP-2025

链接: https://arxiv.org/abs/2510.04655
作者: Yuheng Li,Jiechao Gao,Wei Han,Wenwen Ouyang,Wei Zhu,Hui Yi Leong
机构: Johns Hopkins University (约翰霍普金斯大学); Stanford University (斯坦福大学); Independent Researcher (独立研究者); Carnegie Mellon University (卡内基梅隆大学); University of Hong Kong (香港大学); University of Chicage (芝加哥大学)
类目: Computation and Language (cs.CL)
备注: Accepted by EMNLP-2025 Industrial Track

点击查看摘要

[NLP-42] Evaluating LLM s for Demographic-Targeted Social Bias Detection: A Comprehensive Benchmark Study

链接: https://arxiv.org/abs/2510.04641
作者: Ayan Majumdar,Feihao Chen,Jinghui Li,Xiaozhen Wang
机构: 未知
类目: Computation and Language (cs.CL); Computers and Society (cs.CY); Machine Learning (cs.LG)
备注: 17 pages, 7 figures, 7 tables

点击查看摘要

[NLP-43] Contrastive Learning Using Graph Embeddings for Domain Adaptation of Language Models in the Process Industry EMNLP2025

【速读】：该论文旨在解决工业文本日志中蕴含的结构化知识难以被传统语言模型有效利用的问题，尤其在过程工业领域，这些文本日志通常以稀疏知识图谱（Knowledge Graph, KG）形式存在，但现有方法往往忽略其中的语义关系与领域特定术语。解决方案的关键在于将原本为科学文献设计的图感知邻域对比学习方法（SciNCL）迁移至过程工业场景，通过从GE（工业知识图谱）中提取三元组进行训练，使语言模型能够更好地捕捉文本中的上下文关系与领域知识，从而显著提升文本嵌入质量，在自建的过程工业文本嵌入基准（PITEB）上相较最先进的mE5-large模型性能提升9.8–14.3%（绝对指标提升5.4–8.0p），同时模型参数量减少3–5倍。

链接: https://arxiv.org/abs/2510.04631
作者: Anastasia Zhukova,Jonas Lührs,Christian E. Matt,Bela Gipp
机构: University of Göttingen (哥廷根大学); eschbach GmbH (eschbach公司)
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: accepted to EMNLP 2025 (industry track)

点击查看摘要

Abstract:Recent trends in NLP utilize knowledge graphs (KGs) to enhance pretrained language models by incorporating additional knowledge from the graph structures to learn domain-specific terminology or relationships between documents that might otherwise be overlooked. This paper explores how SciNCL, a graph-aware neighborhood contrastive learning methodology originally designed for scientific publications, can be applied to the process industry domain, where text logs contain crucial information about daily operations and are often structured as sparse KGs. Our experiments demonstrate that language models fine-tuned with triplets derived from GE outperform a state-of-the-art mE5-large text encoder by 9.8-14.3% (5.4-8.0p) on the proprietary process industry text embedding benchmark (PITEB) while being 3-5 times smaller in size.
zh

[NLP-44] Agent ic Context Engineering: Evolving Contexts for Self-Improving Language Models

链接: https://arxiv.org/abs/2510.04618
作者: Qizheng Zhang,Changran Hu,Shubhangi Upasani,Boyuan Ma,Fenglu Hong,Vamsidhar Kamanuru,Jay Rainton,Chen Wu,Mengmeng Ji,Hanchen Li,Urmish Thakker,James Zou,Kunle Olukotun
机构: Stanford University (斯坦福大学); SambaNova Systems, Inc. (桑巴诺瓦系统公司); UC Berkeley (加州大学伯克利分校)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-45] FedSRD: Sparsify-Reconstruct-Decompose for Communication-Efficient Federated Large Language Models Fine-Tuning

链接: https://arxiv.org/abs/2510.04601
作者: Guochen Yan,Luyuan Xie,Qingni Shen,Yuejian Fang,Zhonghai Wu
机构: Peking University (北京大学)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-46] Robustness assessment of large audio language models in multiple-choice evaluation ICASSP2026

链接: https://arxiv.org/abs/2510.04584
作者: Fernando López,Santosh Kesiraju,Jordi Luque
机构: 未知
类目: Computation and Language (cs.CL); Sound (cs.SD); Audio and Speech Processing (eess.AS)
备注: Submitted to ICASSP 2026

点击查看摘要

[NLP-47] Can LLM s Detect Ambiguous Plural Reference? An Analysis of Split-Antecedent and Mereological Reference

【速读】：该论文旨在解决大型语言模型（Large Language Models, LLMs）在歧义与非歧义语境下对复数指代（plural reference）的表征与理解能力问题，具体聚焦于两个核心研究问题：LLMs 是否表现出类人指代偏好，以及能否检测复数回指表达中的歧义并识别可能的指代对象。解决方案的关键在于设计一系列实验，通过下一标记预测任务（next-token prediction tasks）考察代词生成、利用不同提示策略（prompting strategies）评估代词解释与歧义检测能力，并将LLMs的表现与人类行为进行对比分析。结果表明，LLMs在某些情况下能意识到歧义代词的潜在指代对象，但并不总是遵循人类的指代选择，尤其当可选解释未被显式提及；同时，它们在缺乏直接指令时难以识别歧义，且不同实验范式间存在结果不一致性。

链接: https://arxiv.org/abs/2510.04581
作者: Dang Anh,Rick Nouwen,Massimo Poesio
机构: Utrecht University (乌得勒支大学); Queen Mary University of London (伦敦玛丽女王大学)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

Abstract:Our goal is to study how LLMs represent and interpret plural reference in ambiguous and unambiguous contexts. We ask the following research questions: (1) Do LLMs exhibit human-like preferences in representing plural reference? (2) Are LLMs able to detect ambiguity in plural anaphoric expressions and identify possible referents? To address these questions, we design a set of experiments, examining pronoun production using next-token prediction tasks, pronoun interpretation, and ambiguity detection using different prompting strategies. We then assess how comparable LLMs are to humans in formulating and interpreting plural reference. We find that LLMs are sometimes aware of possible referents of ambiguous pronouns. However, they do not always follow human reference when choosing between interpretations, especially when the possible interpretation is not explicitly mentioned. In addition, they struggle to identify ambiguity without direct instruction. Our findings also reveal inconsistencies in the results across different types of experiments.
zh

[NLP-48] LaDiR: Latent Diffusion Enhances LLM s for Text Reasoning

链接: https://arxiv.org/abs/2510.04573
作者: Haoqiang Kang,Yizhe Zhang,Nikki Lijing Kuang,Nicklas Majamaki,Navdeep Jaitly,Yi-An Ma,Lianhui Qin
机构: University of California, San Diego (加州大学圣地亚哥分校); Apple (苹果)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-49] Fine-grained auxiliary learning for real-world product recommendation

链接: https://arxiv.org/abs/2510.04551
作者: Mario Almagro,Diego Ortego,David Jimenez
机构: NielsenIQ(尼尔森IQ)
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: SEPLN 2025

点击查看摘要

[NLP-50] More Than Meets the Eye? Uncovering the Reasoning -Planning Disconnect in Training Vision-Language Driving Models

链接: https://arxiv.org/abs/2510.04532
作者: Xurui Song,Shuo Huai,JingJing Jiang,Jiayi Kong,Jun Luo
机构: S-Lab, Nanyang Technological University, Singapore (南洋理工大学); College of Computing and Data Science, Nanyang Technological University, Singapore (南洋理工大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Robotics (cs.RO)
备注: The dataset will be released publicly once the paper is accepted for publication

点击查看摘要

[NLP-51] ChartAgent : A Multimodal Agent for Visually Grounded Reasoning in Complex Chart Question Answering

链接: https://arxiv.org/abs/2510.04514
作者: Rachneet Kaur,Nishan Srishankar,Zhen Zeng,Sumitra Ganesh,Manuela Veloso
机构: J.P. Morgan AI Research (摩根大通人工智能研究)
类目: Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Methodology (stat.ME)
备注: 53 pages, 12 figures, 15 tables

点击查看摘要

[NLP-52] GRACE: Generative Representation Learning via Contrastive Policy Optimization

链接: https://arxiv.org/abs/2510.04506
作者: Jiashuo Sun,Shixuan Liu,Zhaochen Su,Xianrui Zhong,Pengcheng Jiang,Bowen Jin,Peiran Li,Weijia Shi,Jiawei Han
机构: University of Illinois Urbana–Champaign (伊利诺伊大学厄巴纳-香槟分校); Australian National University (澳大利亚国立大学); Hong Kong University of Science and Technology (香港科技大学); University of Wisconsin–Madison (威斯康星大学麦迪逊分校); University of Washington (华盛顿大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: 23 pages, 7 figures, 7 tables

点击查看摘要

[NLP-53] P2P: A Poison-to-Poison Remedy for Reliable Backdoor Defense in LLM s

【速读】：该论文旨在解决大语言模型（Large Language Models, LLMs）在微调阶段面临的数据投毒后门攻击（data-poisoning backdoor attacks）问题，此类攻击会严重削弱模型的可靠性与可信性。现有防御方法普遍存在泛化能力不足的问题，仅适用于特定攻击类型或任务场景。论文提出的Poison-to-Poison（P2P）算法通过在部分训练样本中注入带有安全替代标签的良性触发词（benign triggers），并利用基于提示的学习（prompt-based learning）对模型进行再微调，使模型将触发词诱导的表征与安全输出关联起来，从而抵消原始恶意触发词的影响。其关键创新在于采用触发词驱动的鲁棒微调机制，实现了跨任务和跨攻击类型的通用防御效果，既有效中和了后门攻击，又保持了原任务性能。

链接: https://arxiv.org/abs/2510.04503
作者: Shuai Zhao,Xinyi Wu,Shiqian Zhao,Xiaobao Wu,Zhongliang Guo,Yanhao Jia,Anh Tuan Luu
机构: Nanyang Technological University (南洋理工大学); Shanghai Jiao Tong University (上海交通大学)
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:

点击查看摘要

Abstract:During fine-tuning, large language models (LLMs) are increasingly vulnerable to data-poisoning backdoor attacks, which compromise their reliability and trustworthiness. However, existing defense strategies suffer from limited generalization: they only work on specific attack types or task settings. In this study, we propose Poison-to-Poison (P2P), a general and effective backdoor defense algorithm. P2P injects benign triggers with safe alternative labels into a subset of training samples and fine-tunes the model on this re-poisoned dataset by leveraging prompt-based learning. This enforces the model to associate trigger-induced representations with safe outputs, thereby overriding the effects of original malicious triggers. Thanks to this robust and generalizable trigger-based fine-tuning, P2P is effective across task settings and attack types. Theoretically and empirically, we show that P2P can neutralize malicious backdoors while preserving task performance. We conduct extensive experiments on classification, mathematical reasoning, and summary generation tasks, involving multiple state-of-the-art LLMs. The results demonstrate that our P2P algorithm significantly reduces the attack success rate compared with baseline models. We hope that the P2P can serve as a guideline for defending against backdoor attacks and foster the development of a secure and trustworthy LLM community.
zh

[NLP-54] GenQuest: An LLM -based Text Adventure Game for Language Learners EMNLP2025

链接: https://arxiv.org/abs/2510.04498
作者: Qiao Wang,Adnan Labib,Robert Swier,Michael Hofmeyr,Zheng Yuan
机构: Hosei University (明治大学); King’s College London (伦敦国王学院); Kindai University (近畿大学); Tokyo Uni. of Science (东京科学大学); University of Sheffield (谢菲尔德大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Workshop on Wordplay: When Language Meets Games, EMNLP 2025

点击查看摘要

[NLP-55] Impatient Users Confuse AI Agents : High-fidelity Simulations of Human Traits for Testing Agents

链接: https://arxiv.org/abs/2510.04491
作者: Muyu He,Anand Kumar,Tsach Mackey,Meghana Rajeev,James Zou,Nazneen Rajani
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 25 pages

点击查看摘要

[NLP-56] Psychological Steering in LLM s: An Evaluation of Effectiveness and Trustworthiness

链接: https://arxiv.org/abs/2510.04484
作者: Amin Banayeeanzade,Ala N. Tak,Fatemeh Bahrani,Anahita Bolourani,Leonardo Blas,Emilio Ferrara,Jonathan Gratch,Sai Praneeth Karimireddy
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Submitted to ARR - October 2025

点击查看摘要

[NLP-57] MedCLM: Learning to Localize and Reason via a CoT-Curriculum in Medical Vision-Language Models

链接: https://arxiv.org/abs/2510.04477
作者: Soo Yong Kim,Suin Cho,Vincent-Daniel Yun,Gyeongyeon Hwang
机构: A.I.MATICS Inc(人工智能矩阵公司); Boston University (波士顿大学); University of Southern California (南加州大学); Heuron(赫罗恩公司); MODULABS, Open Neural Networks Research Lab(模组实验室，开放神经网络研究实验室), OpenAI(开放人工智能公司)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:

点击查看摘要

[NLP-58] Compressed Convolutional Attention: Efficient Attention in a Compressed Latent Space

链接: https://arxiv.org/abs/2510.04476
作者: Tomas Figliolia,Nicholas Alonso,Rishi Iyer,Quentin Anthony,Beren Millidge
机构: Zyphra
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[NLP-59] Mitigating Forgetting Between Supervised and Reinforcement Learning Yields Stronger Reason ers

链接: https://arxiv.org/abs/2510.04454
作者: Xiangchi Yuan,Xiang Chen,Tong Yu,Dachuan Shi,Can Jin,Wenke Lee,Saayan Mitra
机构: Georgia Institute of Technology (佐治亚理工学院); Adobe Research (Adobe 研究院); Rutgers University (罗格斯大学)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-60] On the Role of Unobserved Sequences on Sample-based Uncertainty Quantification for LLM s EMNLP2025

链接: https://arxiv.org/abs/2510.04439
作者: Lucie Kunitomo-Jacquin,Edison Marrese-Taylor,Ken Fukuda
机构: National Institute of Advanced Industrial Science and Technology (AIST), Tokyo, Japan; Graduate School of Engineering, The University of Tokyo, Tokyo, Japan
类目: Computation and Language (cs.CL)
备注: Accepted to UncertaiNLP workshop of EMNLP 2025

点击查看摘要

[NLP-61] Good Intentions Beyond ACL: Who Does NLP for Social Good and Where? EMNLP2025

链接: https://arxiv.org/abs/2510.04434
作者: Grace LeFevre,Qingcheng Zeng,Adam Leif,Jason Jewell,Denis Peskoff,Rob Voigt
机构: Northwestern University (西北大学); University of California, Los Angeles (加州大学洛杉矶分校); University of California, Davis (加州大学戴维斯分校)
类目: Computation and Language (cs.CL); Social and Information Networks (cs.SI)
备注: EMNLP 2025

点击查看摘要

[NLP-62] Partial Information Decomposition via Normalizing Flows in Latent Gaussian Distributions NEURIPS2025

链接: https://arxiv.org/abs/2510.04417
作者: Wenyuan Zhao,Adithya Balachandran,Chao Tian,Paul Pu Liang
机构: Texas A&M University (德州农工大学); Massachusetts Institute of Technology (麻省理工学院)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Information Theory (cs.IT)
备注: NeurIPS 2025

点击查看摘要

[NLP-63] Large Language Models Preserve Semantic Isotopies in Story Continuations

链接: https://arxiv.org/abs/2510.04400
作者: Marc Cavazza
机构: University of Stirling, UK (斯特灵大学); National Institute of Informatics (NII), Tokyo, Japan (日本信息研究所)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[NLP-64] SECA: Semantically Equivalent and Coherent Attacks for Eliciting LLM Hallucinations NEURIPS2025

链接: https://arxiv.org/abs/2510.04398
作者: Buyun Liang,Liangzu Peng,Jinqi Luo,Darshan Thaker,Kwan Ho Ryan Chan,René Vidal
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
备注: Accepted at NeurIPS 2025. Code is available at this https URL

点击查看摘要

[NLP-65] me Is Effort: Estimating Human Post-Editing Time for Grammar Error Correction Tool Evaluation EMNLP2025

链接: https://arxiv.org/abs/2510.04394
作者: Ankit Vadehra,Bill Johnson,Gene Saunders,Pascal Poupart
机构: University of Waterloo (滑铁卢大学), Vector Institute (向量研究所); Scribendi Inc. (斯克里本迪公司)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted for publication in the 4th HCI+NLP Workshop (Fourth Workshop on Bridging Human-Computer Interaction and Natural Language Processing; part of EMNLP 2025)

点击查看摘要

[NLP-66] Improving Consistency in Retrieval-Augmented Systems with Group Similarity Rewards NEURIPS2025

链接: https://arxiv.org/abs/2510.04392
作者: Faisal Hamman,Chenyang Zhu,Anoop Kumar,Xujun Peng,Sanghamitra Dutta,Daben Liu,Alfy Samuel
机构: University of Maryland, College Park (马里兰大学学院公园分校); Capital One (资本one)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Machine Learning (cs.LG)
备注: Accepted at NeurIPS 2025 Workshop on Reliable ML from Unreliable Data

点击查看摘要

[NLP-67] Internal World Models as Imagination Networks in Cognitive Agents

链接: https://arxiv.org/abs/2510.04391
作者: Saurabh Ranjan,Brian Odegaard
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Social and Information Networks (cs.SI); Neurons and Cognition (q-bio.NC)
备注:

点击查看摘要

[NLP-68] MorphoSim: An Interactive Controllable and Editable Language-guided 4D World Simulator

链接: https://arxiv.org/abs/2510.04390
作者: Xuehai He,Shijie Zhou,Thivyanth Venkateswaran,Kaizhi Zheng,Ziyu Wan,Achuta Kadambi,Xin Eric Wang
机构: University of California, Santa Cruz (加州大学圣克鲁兹分校); University of California, Los Angeles (加州大学洛杉矶分校); IIT Bombay (印度理工学院孟买分校); Microsoft (微软)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-69] MacroBench: A Novel Testbed for Web Automation Scripts via Large Language Models NEURIPS2025

链接: https://arxiv.org/abs/2510.04363
作者: Hyunjun Kim,Sejong Kim
机构: KAIST(韩国科学技术院)
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: NeurIPS 2025 Workshop on Lock-LLM

点击查看摘要

[NLP-70] Unmasking Backdoors: An Explainable Defense via Gradient-Attention Anomaly Scoring for Pre-trained Language Models

链接: https://arxiv.org/abs/2510.04347
作者: Anindya Sundar Das,Kangjie Chen,Monowar Bhuyan
机构: Umeå University (于默奥大学); Nanyang Technological University (南洋理工大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 15 pages total (9 pages main text + 4 pages appendix + references), 12 figures, preprint version. The final version may differ

点击查看摘要

[NLP-71] Inoculation Prompting: Eliciting traits from LLM s during training can suppress them at test-time ICLR2026

链接: https://arxiv.org/abs/2510.04340
作者: Daniel Tan,Anders Woodruff,Niels Warncke,Arun Jose,Maxime Riché,David Demitri Africa,Mia Taylor
机构: University College London (伦敦大学学院); Center on Long-Term Risk (长期风险中心); McGill University (麦吉尔大学); Maxime Riché; UK AI Security Institute (英国人工智能安全研究所)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 40 pages, 22 figures In proceedings at ICLR 2026

点击查看摘要

[NLP-72] Evaluation of Clinical Trials Reporting Quality using Large Language Models

链接: https://arxiv.org/abs/2510.04338
作者: Mathieu Laï-king,Patrick Paroubek
机构: Université Paris-Saclay (巴黎萨克雷大学); CNRS (法国国家科学研究中心); LISN (信息与系统科学实验室)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-73] Read the Scene Not the Script: Outcome-Aware Safety for LLM s

链接: https://arxiv.org/abs/2510.04320
作者: Rui Wu,Yihao Quan,Zeru Shi,Zhenting Wang,Yanshu Li,Ruixiang Tang
机构: Rutgers University (罗格斯大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:

点击查看摘要

[NLP-74] Wave-PDE Nets: Trainable Wave-Equation Layers as an Alternative to Attention PRICAI2025

链接: https://arxiv.org/abs/2510.04304
作者: Harshil Vejendla
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: PRICAI 2025 Oral, 9 pages, 3 figures

点击查看摘要

[NLP-75] Measuring Language Model Hallucinations Through Distributional Correctness

链接: https://arxiv.org/abs/2510.04302
作者: Thomas F Burns
机构: Aleph Alpha Research
类目: Computation and Language (cs.CL)
备注: 23 pages, 2 figures

点击查看摘要

[NLP-76] Equipping Retrieval-Augmented Large Language Models with Document Structure Awareness EMNLP2025

链接: https://arxiv.org/abs/2510.04293
作者: Lingnan Xu,Chong Feng,Kaiyuan Zhang,Liu Zhengyong,Wenqiang Xu,Fanqing Meng
机构: Beijing Institute of Technology (北京理工大学); Ant Group (蚂蚁集团)
类目: Computation and Language (cs.CL)
备注: EMNLP2025 Findings

点击查看摘要

[NLP-77] PABSA: Hybrid Framework for Persian Aspect-Based Sentiment Analysis

链接: https://arxiv.org/abs/2510.04291
作者: Mehrzad Tareh,Aydin Mohandesi,Ebrahim Ansari
机构: Institute for Advanced Studies in Basic Sciences (IASBS)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 8 pages

点击查看摘要

[NLP-78] SliceMoE: Routing Embedding Slices Instead of Tokens for Fine-Grained and Balanced Transformer Scaling EMNLP2025

链接: https://arxiv.org/abs/2510.04286
作者: Harshil Vejendla
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: EMNLP 2025 Main, 8 pages, 9 figures

点击查看摘要

[NLP-79] Probing Geometry of Next Token Prediction Using Cumulant Expansion of the Softmax Entropy

链接: https://arxiv.org/abs/2510.04285
作者: Karthik Viswanathan,Sang Eon Park
机构: 未知
类目: Computation and Language (cs.CL); Statistical Mechanics (cond-mat.stat-mech); Machine Learning (cs.LG); Machine Learning (stat.ML)
备注: 14 pages, 7 figures. Poster at HiLD 2025: 3rd Workshop on High-dimensional Learning Dynamics

点击查看摘要

[NLP-80] LongTail-Swap: benchmarking language models abilities on rare words

链接: https://arxiv.org/abs/2510.04268
作者: Robin Algayres,Charles-Éric Saint-James,Mahi Luthra,Jiayi Shen,Dongyan Lin,Youssef Benchekroun,Rashel Moritz,Juan Pino,Emmanuel Dupoux
机构: Meta AI (Meta); EHESS
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[NLP-81] Dont Passmathtt@k: A Bayesian Framework for Large Language Model Evaluation

链接: https://arxiv.org/abs/2510.04265
作者: Mohsen Hariri,Amirhossein Samandar,Michael Hinczewski,Vipin Chaudhary
机构: Case Western Reserve University (凯斯西储大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Statistics Theory (math.ST); Machine Learning (stat.ML)
备注: Code and simulations: this https URL

点击查看摘要

[NLP-82] Pushing on Multilingual Reasoning Models with Language-Mixed Chain-of-Thought

链接: https://arxiv.org/abs/2510.04230
作者: Guijin Son,Donghun Yang,Hitesh Laxmichand Patel,Amit Agarwal,Hyunwoo Ko,Chanuk Lim,Srikant Panda,Minhyuk Kim,Nikunj Drolia,Dasol Choi,Kyong-Ha Lee,Youngjae Yu
机构: 未知
类目: Computation and Language (cs.CL)
备注: Work in Progress

点击查看摘要

[NLP-83] Epistemic Diversity and Knowledge Collapse in Large Language Models

链接: https://arxiv.org/abs/2510.04226
作者: Dustin Wright,Sarah Masud,Jared Moore,Srishti Yadav,Maria Antoniak,Chan Young Park,Isabelle Augenstein
机构: University of Copenhagen (哥本哈根大学); Stanford University (斯坦福大学); University of Colorado Boulder (科罗拉多大学博尔德分校); Microsoft Research (微软研究院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注: 16 pages; 8 figures, 4 tables

点击查看摘要

[NLP-84] Zoom-In to Sort AI-Generated Images Out

链接: https://arxiv.org/abs/2510.04225
作者: Yikun Ji,Yan Hong,Bowen Deng,jun lan,Huijia Zhu,Weiqiang Wang,Liqing Zhang,Jianfu Zhang
机构: Shanghai Jiao Tong University (上海交通大学); Ant Group (蚂蚁集团)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 9 pages, 6 images (19 pages, 11 figures including appendix)

点击查看摘要

[NLP-85] aching LLM to be Persuasive: Reward-Enhanced Policy Optimization for Alignment frm Heterogeneous Rewards

链接: https://arxiv.org/abs/2510.04214
作者: Zhuoran Zhuang,Ye Chen,Xia Zeng,Chao Luo,Luhui Liu,Yihan Chen
机构: Fliggy Alibaba (飞猪阿里巴巴)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-86] CALM Before the STORM: Unlocking Native Reasoning for Optimization Modeling

链接: https://arxiv.org/abs/2510.04204
作者: Zhengyang Tang,Zihan Ye,Chenyu Huang,Xuhan Huang,Chengpeng Li,Sihang Li,Guanhua Chen,Ming Yan,Zizhuo Wang,Hongyuan Zha,Dayiheng Liu,Benyou Wang
机构: The Chinese University of Hong Kong, Shenzhen; Qwen Team, Alibaba Inc.; Shanghai University of Finance and Economics; Southern University of Science and Technology; Shenzhen Loop Area Institute (SLAI)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG)
备注: Work in progress

点击查看摘要

[NLP-87] hinking on the Fly: Test-Time Reasoning Enhancement via Latent Thought Policy Optimization

链接: https://arxiv.org/abs/2510.04182
作者: Wengao Ye,Yan Liang,Lianlei Shan
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[NLP-88] Self Speculative Decoding for Diffusion Large Language Models

链接: https://arxiv.org/abs/2510.04147
作者: Yifeng Gao,Ziang Ji,Yuxuan Wang,Biqing Qi,Hanlin Xu,Linfeng Zhang
机构: Shanghai Jiao Tong University (上海交通大学); University of Science and Technology of China (中国科学技术大学); Xidian University (西安电子科技大学); Shanghai AI Laboratory (上海人工智能实验室); Huawei Technologies Ltd. (华为技术有限公司)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-89] Beyond Next-Token Prediction: A Performance Characterization of Diffusion versus Autoregressive Language Models

链接: https://arxiv.org/abs/2510.04146
作者: Minseo Kim,Coleman Hooper,Aditya Tomar,Chenfeng Xu,Mehrdad Farajtabar,Michael W. Mahoney,Kurt Keutzer,Amir Gholami
机构: Seoul National University (首尔国立大学); University of California, Berkeley (加州大学伯克利分校); ICSI; LBNL; University of Texas at Austin (德克萨斯大学奥斯汀分校); Apple
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 11 pages, 5 figures

点击查看摘要

[NLP-90] Automating construction safety inspections using a multi-modal vision-language RAG framework

链接: https://arxiv.org/abs/2510.04145
作者: Chenxin Wang,Elyas Asadi Shamsabadi,Zhaohui Chen,Luming Shen,Alireza Ahmadian Fard Fini,Daniel Dias-da-Costa
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: 33 pages, 11 figures, 7 tables

点击查看摘要

[NLP-91] Selective Expert Guidance for Effective and Diverse Exploration in Reinforcement Learning of LLM s

链接: https://arxiv.org/abs/2510.04140
作者: Zishang Jiang,Jinyi Han,Tingyun Li,Xinyi Wang,Sihang Jiang,Jiaqing Liang,Zhaoqian Dai,Shuguang Ma,Fei Yu,Yanghua Xiao
机构: Fudan University (复旦大学); East China Normal University (华东师范大学); Ant Group (蚂蚁集团)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-92] Fine Tuning Methods for Low-resource Languages

链接: https://arxiv.org/abs/2510.04139
作者: Tim Bakkenes,Daniel Wang,Anton Johansson
机构: Tsinghua University (清华大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:

点击查看摘要

[NLP-93] Internal states before wait modulate reasoning patterns EMNLP

链接: https://arxiv.org/abs/2510.04128
作者: Dmitrii Troitskii,Koyena Pal,Chris Wendler,Callum Stuart McDougall,Neel Nanda
机构: Northeastern University (东北大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted to EMNLP Findings 2025

点击查看摘要

[NLP-94] Sri Lanka Document Datasets: A Large-Scale Multilingual Resource for Law News and Policy (v20251005)

链接: https://arxiv.org/abs/2510.04124
作者: Nuwan I. Senaratna
机构: Independent Researcher(独立研究员)
类目: Computation and Language (cs.CL)
备注: 4 pages

点击查看摘要

[NLP-95] Unveiling LLM s Metaphorical Understanding: Exploring Conceptual Irrelevance Context Leverag ing and Syntactic Influence

链接: https://arxiv.org/abs/2510.04120
作者: Fengying Ye,Shanshan Wang,Lidia S. Chao,Derek F. Wong
机构: University of Macau (澳门大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[NLP-96] Scaling Code-Assisted Chain-of-Thoughts and Instructions for Model Reasoning NEURIPS2025

链接: https://arxiv.org/abs/2510.04081
作者: Honglin Lin,Qizhi Pei,Xin Gao,Zhuoshi Pan,Yu Li,Juntao Li,Conghui He,Lijun Wu
机构: 未知
类目: Computation and Language (cs.CL); Programming Languages (cs.PL)
备注: Accepted by NeurIPS2025

点击查看摘要

[NLP-97] PoLi-RL: A Point-to-List Reinforcement Learning Framework for Conditional Semantic Textual Similarity

链接: https://arxiv.org/abs/2510.04080
作者: Zixin Song,Bowen Zhang,Qian-Wen Zhang,Di Yin,Xing Sun,Chunping Li
机构: Tsinghua University (清华大学); Tencent Youtu Lab (腾讯优图实验室)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-98] Slow-Fast Policy Optimization: Reposition-Before-Update for LLM Reasoning

链接: https://arxiv.org/abs/2510.04072
作者: Ziyan Wang,Zheng Wang,Jie Fu,Xingwei Qu,Qi Cheng,Shengpu Tang,Minjia Zhang,Xiaoming Huo
机构: Georgia Institute of Technology (佐治亚理工学院); University of Illinois Urbana-Champaign (伊利诺伊大学厄巴纳-香槟分校)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (stat.ML)
备注:

点击查看摘要

[NLP-99] What Makes Diffusion Language Models Super Data Learners?

链接: https://arxiv.org/abs/2510.04071
作者: Zitian Gao,Haoming Luo,Lynx Chen,Jason Klein Liu,Ran Tao,Joey Zhou,Bryan Dai
机构: Ubiquant(万波量化)
类目: Computation and Language (cs.CL)
备注: Technical report, work in progress

点击查看摘要

[NLP-100] What Scales in Cross-Entropy Scaling Law?

链接: https://arxiv.org/abs/2510.04067
作者: Junxi Yan,Zixi Wei,Jingtao Zhan,Qingyao Ai,Yiqun Liu
机构: Tsinghua University (清华大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-101] Exploring Chain-of-Thought Reasoning for Steerable Pluralistic Alignment EMNLP2025 ACL

链接: https://arxiv.org/abs/2510.04045
作者: Yunfan Zhang,Kathleen McKeown,Smaranda Muresan
机构: Columbia University (哥伦比亚大学); Barnard College (巴纳德学院)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: ACL EMNLP 2025

点击查看摘要

[NLP-102] Small Language Models for Emergency Departments Decision Support: A Benchmark Study ATC2025

链接: https://arxiv.org/abs/2510.04032
作者: Zirui Wang,Jiajun Wu,Braden Teitge,Jessalyn Holodinsky,Steve Drew
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted to 2025 IEEE International Conference on Autonomous and Trusted Computing (ATC 2025)

点击查看摘要

[NLP-103] Does Using Counterfactual Help LLM s Explain Textual Importance in Classification?

链接: https://arxiv.org/abs/2510.04031
作者: Nelvin Tan,James Asikin Cheung,Yu-Ching Shih,Dong Yang,Amol Salunkhe
机构: American Express(美国运通)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 8 pages, 2 figures

点击查看摘要

[NLP-104] LLM -Based Data Science Agents : A Survey of Capabilities Challenges and Future Directions

链接: https://arxiv.org/abs/2510.04023
作者: Mizanur Rahman,Amran Bhuiyan,Mohammed Saidul Islam,Md Tahmid Rahman Laskar,Ridwan Mahbub,Ahmed Masry,Shafiq Joty,Enamul Hoque
机构: York University (约克大学); Vector Institute for AI (人工智能研究所); Dialpad Inc. (Dialpad 公司); Nanyang Technological University (南洋理工大学); Salesforce AI Research (Salesforce 人工智能研究)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Survey paper; 45 data science agents; under review

点击查看摘要

[NLP-105] Principled and Tractable RL for Reasoning with Diffusion Language Models

链接: https://arxiv.org/abs/2510.04019
作者: Anthony Zhan
机构: Stanford University (斯坦福大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-106] hai Semantic End-of-Turn Detection for Real-Time Voice Agents ICSE

链接: https://arxiv.org/abs/2510.04016
作者: Thanapol Popit,Natthapath Rungseesiripak,Monthol Charattrakool,Saksorn Ruangtanusak
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: IEEE ICSEC 2025

点击查看摘要

[NLP-107] LLM Microscope: What Model Internals Reveal About Answer Correctness and Context Utilization

链接: https://arxiv.org/abs/2510.04013
作者: Jiarui Liu,Jivitesh Jain,Mona Diab,Nishant Subramani
机构: Carnege Mellon University (卡内基梅隆大学)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-108] Visual Lifelog Retrieval through Captioning-Enhanced Interpretation

链接: https://arxiv.org/abs/2510.04010
作者: Yu-Fei Shih,An-Zi Yen,Hen-Hsen Huang,Hsin-Hsi Chen
机构: 未知
类目: Information Retrieval (cs.IR); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
备注:

点击查看摘要

[NLP-109] What Shapes a Creative Machine Mind? Comprehensively Benchmarking Creativity in Foundation Models

链接: https://arxiv.org/abs/2510.04009
作者: Zicong He,Boxuan Zhang,Weihao Liu,Ruixiang Tang,Lu Cheng
机构: Georgia Institute of Technology (佐治亚理工学院); Rutgers University (罗格斯大学); University of Illinois Chicago (芝加哥伊利诺伊大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 22 pages

点击查看摘要

[NLP-110] Enhancing OCR for Sino-Vietnamese Language Processing via Fine-tuned PaddleOCRv5

链接: https://arxiv.org/abs/2510.04003
作者: Minh Hoang Nguyen,Su Nguyen Thiet
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: 5 pages, 6 figures, 2 tables

点击查看摘要

[NLP-111] AgriGPT -VL: Agricultural Vision-Language Understanding Suite

链接: https://arxiv.org/abs/2510.04002
作者: Bo Yang,Yunkui Chen,Lanfei Feng,Yu Zhang,Xiao Xu,Jianyu Zhang,Nueraili Aierken,Runhe Huang,Hongjian Lin,Yibin Ying,Shijian Li
机构: Zhejiang University (浙江大学); Hosei University (法政大学)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-112] Named Entity Recognition in COVID-19 tweets with Entity Knowledge Augmentation

链接: https://arxiv.org/abs/2510.04001
作者: Xuankang Zhang,Jiangming Liu
机构: Yunnan University (云南大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Work in progress

点击查看摘要

[NLP-113] Simulating and Understanding Deceptive Behaviors in Long-Horizon Interactions

链接: https://arxiv.org/abs/2510.03999
作者: Yang Xu,Xuanming Zhang,Min-Hsuan Yeh,Jwala Dhamala,Ousmane Dia,Rahul Gupta,Yixuan Li
机构: University of Wisconsin-Madison (威斯康星大学麦迪逊分校); Zhejiang University (浙江大学); Amazon (亚马逊)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-114] Mapping Patient-Perceived Physician Traits from Nationwide Online Reviews with LLM s

【速读】：该论文旨在解决如何通过大规模患者文本数据自动量化患者对医生的主观感知问题，以提升医患信任、沟通与满意度。其解决方案的关键在于构建了一个基于大语言模型（Large Language Model, LLM）的分析管道，能够从410万条美国医生的患者评论中推断出五大性格特质（Big Five personality traits）及五项以患者为中心的主观判断，并通过多模型比较和专家人工标注验证了方法的有效性，相关系数达0.72–0.89，且与患者满意度显著正相关（r = 0.41–0.81, p < 0.001），从而实现了可解释、可验证的大规模医患关系分析。

链接: https://arxiv.org/abs/2510.03997
作者: Junjie Luo,Rui Han,Arshana Welivita,Zeleikun Di,Jingfu Wu,Xuzhe Zhi,Ritu Agarwal,Gordon Gao
机构: Johns Hopkins School of Medicine (约翰霍普金斯医学院); Johns Hopkins University (约翰霍普金斯大学); Carey Business School, Johns Hopkins University (约翰霍普金斯大学凯瑞商学院); Center for Digital Health Artificial Intelligence (CDHAI) (数字健康人工智能中心)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

Abstract:Understanding how patients perceive their physicians is essential to improving trust, communication, and satisfaction. We present a large language model (LLM)-based pipeline that infers Big Five personality traits and five patient-oriented subjective judgments. The analysis encompasses 4.1 million patient reviews of 226,999 U.S. physicians from an initial pool of one million. We validate the method through multi-model comparison and human expert benchmarking, achieving strong agreement between human and LLM assessments (correlation coefficients 0.72-0.89) and external validity through correlations with patient satisfaction (r = 0.41-0.81, all p0.001). National-scale analysis reveals systematic patterns: male physicians receive higher ratings across all traits, with largest disparities in clinical competence perceptions; empathy-related traits predominate in pediatrics and psychiatry; and all traits positively predict overall satisfaction. Cluster analysis identifies four distinct physician archetypes, from “Well-Rounded Excellent” (33.8%, uniformly high traits) to “Underperforming” (22.6%, consistently low). These findings demonstrate that automated trait extraction from patient narratives can provide interpretable, validated metrics for understanding physician-patient relationships at scale, with implications for quality measurement, bias detection, and workforce development in healthcare.
zh

[NLP-115] No Tokens Wasted: Leverag ing Long Context in Biomedical Vision-Language Models

链接: https://arxiv.org/abs/2510.03978
作者: Min Woo Sun,Alejandro Lozano,Javier Gamazo Tejero,Vishwesh Nath,Xiao Xiao Sun,James Burgess,Yuhui Zhang,Kun Yuan,Robert Tibshirani,Sean Huver,Serena Yeung-Levy
机构: Stanford University (斯坦福大学); NVIDIA (英伟达)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-116] LLM Chemistry Estimation for Multi-LLM Recommendation

链接: https://arxiv.org/abs/2510.03930
作者: Huascar Sanchez,Briland Hitaj
机构: SRI International (SRI国际)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 20 pages, 5 figures, 5 tables

点击查看摘要

[NLP-117] PsycholexTherapy: Simulating Reasoning in Psychotherapy with Small Language Models in Persian

链接: https://arxiv.org/abs/2510.03913
作者: Mohammad Amin Abbasi,Hassan Naderi
机构: 未知
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-118] Read Between the Lines: A Benchmark for Uncovering Political Bias in Bangla News Articles

链接: https://arxiv.org/abs/2510.03898
作者: Nusrat Jahan Lia,Shubhashis Roy Dipta,Abdullah Khan Zehady,Naymul Islam,Madhusodan Chakraborty,Abdullah Al Wasif
机构: University of Dhaka (达卡大学); University of Maryland, Baltimore County (马里兰大学巴尔的摩县分校); Cisco Systems (思科系统公司); BanglaLLM; Maharishi International University (玛赫西国际大学); Unityflow AI
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-119] Kantian-Utilitarian XAI: Meta-Explained

链接: https://arxiv.org/abs/2510.03892
作者: Zahra Atf,Peter R. Lewis
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted for presentation as a poster at the 35th IEEE International Conference on Collaborative Advances in Software and Computing, 2025. Conference website: this https URL

点击查看摘要

[NLP-120] Unlocking Reasoning Capabilities in LLM s via Reinforcement Learning Exploration

链接: https://arxiv.org/abs/2510.03865
作者: Wenhao Deng,Long Wei,Chenglei Yu,Tailin Wu
机构: Westlake University (西湖大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-121] Annotate Rhetorical Relations with INCEpTION: A Comparison with Automatic Approaches

链接: https://arxiv.org/abs/2510.03808
作者: Mehedi Hasan Emon
机构: 未知
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-122] Beyond Token Length: Step Pruner for Efficient and Accurate Reasoning in Large Language Models

链接: https://arxiv.org/abs/2510.03805
作者: Canhui Wu,Qiong Cao,Chang Li,Zhenfang Wang,Chao Xue,Yuwei Fan,Wei Xi,Xiaodong He
机构: Xi’an Jiaotong University (西安交通大学); JD Future Academy (京东未来研究院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 20pages, 7 figures

点击查看摘要

[NLP-123] Mechanistic Interpretability of Socio-Political Frames in Language Models ECML KDD2024

链接: https://arxiv.org/abs/2510.03799
作者: Hadi Asghari,Sami Nenno
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注: Peer-reviewed and presented at Advances in Interpretable Machine Learning and Artificial Intelligence (AIMLAI) Workshop at ECML/PKDD 2024

点击查看摘要

[NLP-124] Investigating LLM Variability in Personalized Conversational Information Retrieval SIGIR

链接: https://arxiv.org/abs/2510.03795
作者: Simon Lupart,Daniël van Dijk,Eric Langezaal,Ian van Dort,Mohammad Aliannejadi
机构: University of Amsterdam (阿姆斯特丹大学)
类目: Information Retrieval (cs.IR); Computation and Language (cs.CL)
备注: 11 pages, 5 figures, SIGIR-AP’25 Proceedings of the 2025 Annual International ACM SIGIR Conference on Research and Development in Information Retrieval in the Asia Pacific Region (SIGIR-AP 2025), December 7–10, 2025, Xi’an, China

点击查看摘要

[NLP-125] Rezwan: Leverag ing Large Language Models for Comprehensive Hadith Text Processing: A 1.2M Corpus Development

链接: https://arxiv.org/abs/2510.03781
作者: Majid Asgari-Bidhendi,Muhammad Amin Ghaseminia,Alireza Shahbazi,Sayyed Ali Hossayni,Najmeh Torabian,Behrouz Minaei-Bidgoli
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 9 pages, 3 figures

点击查看摘要

[NLP-126] Prompt Balance Matters: Understanding How Imbalanced Few-Shot Learning Affects Multilingual Sense Disambiguation in LLM s

链接: https://arxiv.org/abs/2510.03762
作者: Deshan Sumanathilaka,Nicholas Micallef,Julian Hough
机构: Swansea University (斯旺西大学)
类目: Computation and Language (cs.CL)
备注: Paper accepted at GlobalNLP 2025: Workshop on beyond English: Natural Language Processing for All Languages in an Era of Large Language Models" 9 pages, 3 figures, 2 Tables

点击查看摘要

[NLP-127] Cross-Lingual Multi-Granularity Framework for Interpretable Parkinsons Disease Diagnosis from Speech

链接: https://arxiv.org/abs/2510.03758
作者: Ilias Tougui,Mehdi Zakroum,Mounir Ghogho
机构: 未知
类目: Computation and Language (cs.CL); Sound (cs.SD); Audio and Speech Processing (eess.AS)
备注:

点击查看摘要

[NLP-128] reePrompt: Leverag ing Hierarchical Few-Shot Example Selection for Improved English-Persian and English-German Translation

链接: https://arxiv.org/abs/2510.03748
作者: Ramtin Kakavand,Ebrahim Ansari
机构: Institute for Advanced Studies in Basic Sciences (基础科学高级研究所)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 12 pages

点击查看摘要

[NLP-129] Optimizing Fine-Tuning through Advanced Initialization Strategies for Low-Rank Adaptation

链接: https://arxiv.org/abs/2510.03731
作者: Yongfu Xue
机构: Tongji University (同济大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-130] Bridging the Gap Between Multimodal Foundation Models and World Models

链接: https://arxiv.org/abs/2510.03727
作者: Xuehai He
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: PhD thesis

点击查看摘要

[NLP-131] Person-Centric Annotations of LAION-400M: Auditing Bias and Its Transfer to Models

链接: https://arxiv.org/abs/2510.03721
作者: Leander Girrbach,Stephan Alaniz,Genevieve Smith,Trevor Darrell,Zeynep Akata
机构: Technical University of Munich (慕尼黑工业大学); Munich Center for Machine Learning (慕尼黑机器学习中心); MDSI; LTCI (电信信息实验室); Télécom Paris (巴黎电信学院); Institut Polytechnique de Paris (巴黎综合理工学院); Helmholtz Munich (赫尔姆霍兹慕尼黑研究中心); University of California, Berkeley (加州大学伯克利分校)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Computers and Society (cs.CY); Machine Learning (cs.LG)
备注: 48 pages

点击查看摘要

[NLP-132] Mind the Goal: Data-Efficient Goal-Oriented Evaluation of Conversational Agents and Chatbots using Teacher Models

链接: https://arxiv.org/abs/2510.03696
作者: Deepak Babu Piskala,Sharlene Chen,Udita Patel,Parul Kalra,Rafael Castrillo
机构: Amazon.com(亚马逊)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-133] MedReflect: Teaching Medical LLM s to Self-Improve via Reflective Correction

链接: https://arxiv.org/abs/2510.03687
作者: Yue Huang,Yanyuan Chen,Dexuan Xu,Weihua Yue,Huamin Zhang,Meikang Qiu,Yu Huang
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[NLP-134] Fine-Tuning Large Language Models with QLoRA for Offensive Language Detection in Roman Urdu-English Code-Mixed Text

链接: https://arxiv.org/abs/2510.03683
作者: Nisar Hussain,Amna Qasim,Gull Mehak,Muhammad Zain,Momina Hafeez,Grigori Sidorov
机构: 未知
类目: Computation and Language (cs.CL)
备注: 25 pages, 22 figures

点击查看摘要

[NLP-135] oken Hidden Reward: Steering Exploration-Exploitation in Group Relative Deep Reinforcement Learning

链接: https://arxiv.org/abs/2510.03669
作者: Wenlong Deng,Yi Ren,Yushu Li,Boying Gong,Danica J. Sutherland,Xiaoxiao Li,Christos Thrampoulidis
机构: University of British Columbia (不列颠哥伦比亚大学); Vector Institute (矢量研究所); Amii (阿米); Meta (Meta)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-136] UNIDOC-BENCH: A Unified Benchmark for Document-Centric Multimodal RAG

链接: https://arxiv.org/abs/2510.03663
作者: Xiangyu Peng,Cab Qin,Zeyuan Chen,Ran Xu,Caiming Xiong,Chien-Sheng Wu
机构: Salesforce AI Research (Salesforce人工智能研究)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-137] Does higher interpretability imply better utility? A Pairwise Analysis on Sparse Autoencoders

链接: https://arxiv.org/abs/2510.03659
作者: Xu Wang,Yan Hu,Benyou Wang,Difan Zou
机构: The University of Hong Kong (香港大学); The Chinese University of Hong Kong, Shenzhen (深圳清华大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (stat.ML)
备注: 24 pages

点击查看摘要

[NLP-138] owards Unsupervised Speech Recognition at the Syllable-Level

链接: https://arxiv.org/abs/2510.03639
作者: Liming Wang,Junrui Ni,Kai-Wei Chang,Saurabhchand Bhati,David Harwath,Mark Hasegawa-Johnson,James R. Glass
机构: Massachusetts Institute of Technology (麻省理工学院); University of Illinois Urbana-Champaign (伊利诺伊大学厄巴纳-香槟分校); University of Texas at Austin (德克萨斯大学奥斯汀分校); Amazon (亚马逊)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[NLP-139] From Theory to Practice: Evaluating Data Poisoning Attacks and Defenses in In-Context Learning on Social Media Health Discourse

链接: https://arxiv.org/abs/2510.03636
作者: Rabeya Amin Jhuma,Mostafa Mohaimen Akand Faisal
机构: University of Information Technology and Sciences (UITS)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Cryptography and Security (cs.CR)
备注:

点击查看摘要

[NLP-140] Can an LLM Induce a Graph? Investigating Memory Drift and Context Length

链接: https://arxiv.org/abs/2510.03611
作者: Raquib Bin Yousuf,Aadyant Khatri,Shengzhe Xu,Mandar Sharma,Naren Ramakrishnan
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 2025 IEEE International Conference on Knowledge Graph (ICKG)

点击查看摘要

[NLP-141] Decoupling Task-Solving and Output Formatting in LLM Generation

链接: https://arxiv.org/abs/2510.03595
作者: Haikang Deng,Po-Nien Kung,Nanyun Peng
机构: University of California, Los Angeles (加州大学洛杉矶分校)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-142] LLM Reporting In! Medical Information Extraction Across Prompting Fine-tuning and Post-correction

链接: https://arxiv.org/abs/2510.03577
作者: Ikram Belmadani,Parisa Nazari Hashemi,Thomas Sebbag,Benoit Favre,Guillaume Fortier,Solen Quiniou,Emmanuel Morin,Richard Dufour
机构: Nantes Université, École Centrale Nantes, CNRS, LS2N, UMR 6004, F-44000 Nantes, France; Aix-Marseille Université, CNRS, LIS UMR 7020, 13000, Marseille, France; Explore, Carquefou, France; Inetum, 93400 Saint-Ouen-sur-Seine, France
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: in French language

点击查看摘要

[NLP-143] Machine Unlearning Meets Adversarial Robustness via Constrained Interventions on LLM s

链接: https://arxiv.org/abs/2510.03567
作者: Fatmazohra Rezkellah,Ramzi Dakhmouche
机构: Université Paris-Dauphine (巴黎第九大学); EPFL (瑞士联邦理工学院); Empa (瑞士联邦材料科学与技术研究所)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Cryptography and Security (cs.CR); Computers and Society (cs.CY); Optimization and Control (math.OC)
备注:

点击查看摘要

[NLP-144] Reactive Transformer (RxT) – Stateful Real-Time Processing for Event-Driven Reactive Language Models

链接: https://arxiv.org/abs/2510.03561
作者: Adam Filipek
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 25 pages, 13 figures

点击查看摘要

[NLP-145] CCD-Bench: Probing Cultural Conflict in Large Language Model Decision-Making

链接: https://arxiv.org/abs/2510.03553
作者: Hasibur Rahman,Hanan Salam
机构: 未知
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-146] What is a protest anyway? Codebook conceptualization is still a first-order concern in LLM -era classification

链接: https://arxiv.org/abs/2510.03541
作者: Andrew Halterman,Katherine A. Keith
机构: Michigan State University (密歇根州立大学); Williams College (威廉姆斯学院)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-147] riMediQ: A Triplet-Structured Approach for Interactive Medical Question Answering

链接: https://arxiv.org/abs/2510.03536
作者: Zhaohan Meng,Zaiqiao Meng,Siwei Liu,Iadh Ounis
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Preprint

点击查看摘要

[NLP-148] Fine-Tuning on Noisy Instructions: Effects on Generalization and Performance

链接: https://arxiv.org/abs/2510.03528
作者: Ahmed Alajrami,Xingwei Tan,Nikolaos Aletras
机构: University of Sheffield (谢菲尔德大学)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-149] Sample Align Synthesize: Graph-Based Response Synthesis with ConGrs

链接: https://arxiv.org/abs/2510.03527
作者: Sayan Ghosh,Shahzaib Saqib Warraich,Dhruv Tarsadiya,Gregory Yauney,Swabha Swayamdipta
机构: University of Southern California (南加州大学)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-150] Identifying Financial Risk Information Using RAG with a Contrastive Insight NEURIPS2025

链接: https://arxiv.org/abs/2510.03521
作者: Ali Elahi
机构: University of Illinois Chicago (伊利诺伊大学芝加哥分校); Surlamer Investments (苏拉默投资公司)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 7 pages, 1 figure, Workshop on Generative AI in Finance, NeurIPS 2025

点击查看摘要

[NLP-151] S-Reason er: Aligning Time Series Foundation Models with LLM Reasoning

链接: https://arxiv.org/abs/2510.03519
作者: Fangxu Yu,Hongyu Zhao,Tianyi Zhou
机构: University of Maryland, College Park (马里兰大学学院市分校)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[NLP-152] Red Lines and Grey Zones in the Fog of War: Benchmarking Legal Risk Moral Harm and Regional Bias in Large Language Model Military Decision-Making

链接: https://arxiv.org/abs/2510.03514
作者: Toby Drinkall
机构: 未知
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 54 pages; 11 figures

点击查看摘要

[NLP-153] ALHD: A Large-Scale and Multigenre Benchmark Dataset for Arabic LLM -Generated Text Detection

链接: https://arxiv.org/abs/2510.03502
作者: Ali Khairallah,Arkaitz Zubiaga
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 47 pages, 15 figures. Dataset available at Zenodo: this https URL Codebase available at GitHub: this https URL

点击查看摘要

[NLP-154] SEER: The Span-based Emotion Evidence Retrieval Benchmark

链接: https://arxiv.org/abs/2510.03490
作者: Aneesha Sampath,Oya Aran,Emily Mower Provost
机构: University of Michigan (密歇根大学); Procter & Gamble (宝洁公司)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[NLP-155] Searching for the Most Human-like Emergent Language

【速读】：该论文旨在解决生成式 AI (Generative AI) 中涌现语言（emergent language）与人类语言之间相似性不足的问题，即如何设计出更接近人类语言特征的通信协议。其解决方案的关键在于构建一个基于信号博弈（signalling game）的环境，并通过超参数优化来提升涌现语言的统计特性，其中以 XferBench 作为目标函数，量化其对深度迁移学习（deep transfer learning）到人类语言的适用性；同时发现熵（entropy）具有预测涌现语言迁移性能的能力，并验证了涌现通信系统中熵最小化的特性，从而为生成更具现实性的涌现语言提供了可泛化的超参数配置依据。

链接: https://arxiv.org/abs/2510.03467
作者: Brendon Boldt,David Mortensen
机构: Carnegie Mellon University (卡内基梅隆大学)
类目: Computation and Language (cs.CL)
备注: Accepted for publication at the 2025 Conference on Empirical Methods in Natural Language Processing; 19 pages, 12 figures

点击查看摘要

Abstract:In this paper, we design a signalling game-based emergent communication environment to generate state-of-the-art emergent languages in terms of similarity to human language. This is done with hyperparameter optimization, using XferBench as the objective function. XferBench quantifies the statistical similarity of emergent language to human language by measuring its suitability for deep transfer learning to human language. Additionally, we demonstrate the predictive power of entropy on the transfer learning performance of emergent language as well as corroborate previous results on the entropy-minimization properties of emergent communication systems. Finally, we report generalizations regarding what hyperparameters produce more realistic emergent languages, that is, ones which transfer better to human language.
zh

[NLP-156] Omni-Embed-Nemotron: A Unified Multimodal Retrieval Model for Text Image Audio and Video

链接: https://arxiv.org/abs/2510.03458
作者: Mengyao Xu,Wenfei Zhou,Yauhen Babakhin,Gabriel Moreira,Ronay Ak,Radek Osmulski,Bo Liu,Even Oldridge,Benedikt Schifferer
机构: NVIDIA(英伟达); NVIDIA(英伟达); NVIDIA(英伟达); NVIDIA(英伟达); NVIDIA(英伟达); NVIDIA(英伟达); NVIDIA(英伟达); NVIDIA(英伟达)
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-157] Morpheme Induction for Emergent Language

链接: https://arxiv.org/abs/2510.03439
作者: Brendon Boldt,David Mortensen
机构: Carnegie Mellon University (卡内基梅隆大学)
类目: Computation and Language (cs.CL)
备注: Accepted for publication at the 2025 Conference on Empirical Methods in Natural Language Processing; 16 pages, 4 figures

点击查看摘要

[NLP-158] Consistent Kernel Change-Point Detection under m-Dependence for Text Segmentation

链接: https://arxiv.org/abs/2510.03437
作者: Jairo Diaz-Rodriguez,Mumin Jia
机构: York University (约克大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Machine Learning (stat.ML)
备注:

点击查看摘要

[NLP-159] PLSEMANTICSBENCH: Large Language Models As Programming Language Interpreters

链接: https://arxiv.org/abs/2510.03415
作者: Aditya Thimmaiah,Jiyang Zhang,Jayanth Srinivasa,Junyi Jessy Li,Milos Gligoric
机构: The University of Texas at Austin (德克萨斯大学奥斯汀分校); Cisco Research (思科研究院)
类目: Programming Languages (cs.PL); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Software Engineering (cs.SE)
备注:

点击查看摘要

[NLP-160] Know Thyself? On the Incapability and Implications of AI Self-Recognition

链接: https://arxiv.org/abs/2510.03399
作者: Xiaoyan Bai,Aryan Shrivastava,Ari Holtzman,Chenhao Tan
机构: University of Chicago (芝加哥大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computers and Society (cs.CY); Machine Learning (cs.LG)
备注: Our code is available, see this https URL

点击查看摘要

[NLP-161] Studying the Korean Word-Chain Game with RLVR:Mitigating Reward Conflicts via Curriculum Learning

链接: https://arxiv.org/abs/2510.03394
作者: Donghwan Rho
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 10 pages

点击查看摘要

[NLP-162] Implicit Values Embedded in How Humans and LLM s Complete Subjective Everyday Tasks

链接: https://arxiv.org/abs/2510.03384
作者: Arjun Arunasalam,Madison Pickering,Z. Berkay Celik,Blase Ur
机构: Florida International University (佛罗里达国际大学); University of Chicago (芝加哥大学); Purdue University (普渡大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[NLP-163] Lightweight Prompt Engineering for Cognitive Alignment in Educational AI: A OneClickQuiz Case Study CEC

链接: https://arxiv.org/abs/2510.03374
作者: Antoun Yaacoub,Zainab Assaghir,Jérôme Da-Rugna
机构: 未知
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Published in the 36th Central European Conference on Information and Intelligent Systems(CECIIS)at: Varaždin, Croatia. September 17-19/2025. ISSN 1847-2001 (Print). ISSN 1848-2295 (Online)

点击查看摘要

[NLP-164] Agent Caster: Reasoning -Guided Tornado Forecasting

链接: https://arxiv.org/abs/2510.03349
作者: Michael Chen
机构: California Institute of Technology (加州理工学院)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Atmospheric and Oceanic Physics (physics.ao-ph)
备注:

点击查看摘要

[NLP-165] Graph-S3: Enhancing Agent ic textual Graph Retrieval with Synthetic Stepwise Supervision

链接: https://arxiv.org/abs/2510.03323
作者: Ge Chang,Jinbo Su,Jiacheng Liu,Pengfei Yang,Yuhao Shang,Huiwen Zheng,Hongli Ma,Yan Liang,Yuanchun Li,Yunxin Liu
机构: Institute for AI Industry Research (AIR), Tsinghua University; Beijing Academy of Artificial Intelligence (BAAI); GDS Holdings Limited
类目: Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-166] Decomposing Attention To Find Context-Sensitive Neurons NEURIPS2025

链接: https://arxiv.org/abs/2510.03315
作者: Alex Gibson
机构: University of Cambridge (剑桥大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 10 pages, 7 figures. Submitted to the Mechanistic Interpretability Workshop at NeurIPS 2025

点击查看摘要

[NLP-167] CAFL-L: Constraint-Aware Federated Learning with Lagrangian Dual Optimization for On-Device Language Models NEURIPS

链接: https://arxiv.org/abs/2510.03298
作者: Dongqi Zheng,Wenjin Fu
机构: Purdue University (普渡大学); Apple Inc. (苹果公司); Carnegie Mellon University (卡内基梅隆大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Distributed, Parallel, and Cluster Computing (cs.DC)
备注: Accepted by 39th NeurIPS - Constrained Optimization for Machine Learning

点击查看摘要

[NLP-168] Multimodal Arabic Captioning with Interpretable Visual Concept Integration

链接: https://arxiv.org/abs/2510.03295
作者: Passant Elchafei,Amany Fashwan
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:

点击查看摘要

[NLP-169] Why mask diffusion does not work

链接: https://arxiv.org/abs/2510.03289
作者: Haocheng Sun,Cynthia Xin Wen,Edward Hong Wang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-170] MACE: A Hybrid LLM Serving System with Colocated SLO-aware Continuous Retraining Alignment

链接: https://arxiv.org/abs/2510.03283
作者: Yufei Li,Yu Fu,Yue Dong,Cong Liu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Distributed, Parallel, and Cluster Computing (cs.DC)
备注: 14 pages, 15 figures

点击查看摘要

[NLP-171] Discovering Transformer Circuits via a Hybrid Attribution and Pruning Framework NEURIPS2025

【速读】：该论文旨在解决语言模型可解释性研究中电路发现（circuit discovery）算法的效率与忠实性之间的权衡问题：现有方法中，归因打补丁（attribution patching）虽速度快但对完整模型的忠实度低，而边剪枝（edge pruning）虽忠实但计算成本高。解决方案的关键在于提出一种混合归因与剪枝（Hybrid Attribution and Pruning, HAP）框架，首先利用归因打补丁快速定位高潜力子图，再在此基础上应用边剪枝提取忠实的稀疏电路，从而在保持电路忠实性的同时提升计算效率达46%。

链接: https://arxiv.org/abs/2510.03282
作者: Hao Gu,Vibhas Nair,Amrithaa Ashok Kumar,Jayvart Sharma,Ryan Lagasse
机构: Algoverse AI Research
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Accepted to the NeurIPS 2025 Workshop on Mechanistic Interpretability (Mechinterp) and the NeurIPS 2025 Workshop on New Perspectives in Graph Machine Learning

点击查看摘要

Abstract:Interpreting language models often involves circuit analysis, which aims to identify sparse subnetworks, or circuits, that accomplish specific tasks. Existing circuit discovery algorithms face a fundamental trade-off: attribution patching is fast but unfaithful to the full model, while edge pruning is faithful but computationally expensive. This research proposes a hybrid attribution and pruning (HAP) framework that uses attribution patching to identify a high-potential subgraph, then applies edge pruning to extract a faithful circuit from it. We show that HAP is 46% faster than baseline algorithms without sacrificing circuit faithfulness. Furthermore, we present a case study on the Indirect Object Identification task, showing that our method preserves cooperative circuit components (e.g. S-inhibition heads) that attribution patching methods prune at high sparsity. Our results show that HAP could be an effective approach for improving the scalability of mechanistic interpretability research to larger models. Our code is available at this https URL.
zh

[NLP-172] raining Optimal Large Diffusion Language Models

链接: https://arxiv.org/abs/2510.03280
作者: Jinjie Ni,Qian Liu,Chao Du,Longxu Dou,Hang Yan,Zili Wang,Tianyu Pang,Michael Qizhe Shieh
机构: National University of Singapore(新加坡国立大学); Sea AI Lab; StepFun; Shanghai Qiji Zhifeng Co., Ltd.(上海奇绩创坛有限公司)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-173] MemMamba: Rethinking Memory Patterns in State Space Model

链接: https://arxiv.org/abs/2510.03279
作者: Youjin Wang,Yangjingyi Chen,Jiahao Yan,Jiaxuan Lu,Xiao Sun
机构: Renmin University of China (中国人民大学); Shanghai University of Finance and Economics (上海财经大学); Gao Ling Institute of Artificial Intelligence (高瓴人工智能学院); Shanghai Artificial Intelligence Laboratory (上海人工智能实验室)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-174] General Exploratory Bonus for Optimistic Exploration in RLHF

【速读】：该论文旨在解决强化学习与人类反馈（Reinforcement Learning with Human Feedback, RLHF）中探索效率低下的问题，特别是现有探索奖励（exploratory bonus）方法在KL散度或α-散度正则化下无法实现乐观探索（optimistic exploration）的缺陷——这些方法会无意中将探索引导至参考模型高概率区域，导致保守行为而非对不确定区域的主动发现。解决方案的关键在于提出一种全新的理论框架——通用探索奖励（General Exploratory Bonus, GEB），其通过参考依赖的奖励调节机制有效抵消由散度诱导的偏差，并严格满足乐观原则；GEB不仅可统一此前启发式奖励为特例，还能自然扩展至整个α-散度族，在多个任务和大语言模型基座上均显著优于基线方法。

链接: https://arxiv.org/abs/2510.03269
作者: Wendi Li,Changdae Oh,Yixuan Li
机构: University of Wisconsin–Madison (威斯康星大学麦迪逊分校)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:

点击查看摘要

Abstract:Optimistic exploration is central to improving sample efficiency in reinforcement learning with human feedback, yet existing exploratory bonus methods to incentivize exploration often fail to realize optimism. We provide a theoretical analysis showing that current formulations, under KL or \alpha -divergence regularization, unintentionally bias exploration toward high-probability regions of the reference model, thereby reinforcing conservative behavior instead of promoting discovery of uncertain regions. To address this pitfall, we introduce the General Exploratory Bonus (GEB), a novel theoretical framework that provably satisfies the optimism principle. GEB counteracts divergence-induced bias via reference-dependent reward regulation and unifies prior heuristic bonuses as special cases, while extending naturally across the full \alpha -divergence family. Empirically, GEB consistently outperforms baselines on alignment tasks across multiple divergence settings and large language model backbones. These results demonstrate that GEB offers both a principled and practical solution for optimistic exploration in RLHF.
zh

[NLP-175] Share Your Attention: Transformer Weight Sharing via Matrix-based Dictionary Learning

链接: https://arxiv.org/abs/2508.04581
作者: Magauiya Zhussip,Dmitriy Shopkhoev,Ammar Ali,Stamatios Lefkimmiatis
机构: MTS AI; ITMO University
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[NLP-176] ReplaceMe: Network Simplification via Depth Pruning and Transformer Block Linearization

链接: https://arxiv.org/abs/2505.02819
作者: Dmitriy Shopkhoev,Ammar Ali,Magauiya Zhussip,Valentin Malykh,Stamatios Lefkimmiatis,Nikos Komodakis,Sergey Zagoruyko
机构: MTS AI; ITMO University; IITU; University of Crete; IACM-Forth; Archimedes Athena RC; Polynome
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[NLP-177] CAG: Chunked Augmented Generation for Google Chromes Built-in Gemini Nano

链接: https://arxiv.org/abs/2412.18708
作者: Vivek Vellaiyappan Surulimuthu,Aditya Karnam Gururaj Rao
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC); Information Retrieval (cs.IR)
备注: 36 pages, 19 figures

点击查看摘要

[NLP-178] Large Language Models Achieve Gold Medal Performance at International Astronomy Astrophysics Olympiad

链接: https://arxiv.org/abs/2510.05016
作者: Lucas Carrit Delgado Pinheiro,Ziru Chen,Bruno Caixeta Piazza,Ness Shroff,Yingbin Liang,Yuan-Sen Ting,Huan Sun
机构: 未知
类目: Instrumentation and Methods for Astrophysics (astro-ph.IM); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 18 pages, 6 figures, to be submitted, comments are welcome

点击查看摘要

[NLP-179] AtomWorld: A Benchmark for Evaluating Spatial Reasoning in Large Language Models on Crystalline Materials

链接: https://arxiv.org/abs/2510.04704
作者: Taoyuze Lv,Alexander Chen,Fengyu Xie,Chu Wu,Jeffrey Meng,Dongzhan Zhou,Bram Hoex,Zhicheng Zhong,Tong Xie
机构: Suzhou Institute for Advanced Research (苏州研究院); University of Science and Technology of China (中国科学技术大学); University of New South Wales (新南威尔士大学); Shanghai Artificial Intelligence Laboratory (上海人工智能实验室)
类目: Materials Science (cond-mat.mtrl-sci); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:

点击查看摘要

[NLP-180] Adapting Diarization-Conditioned Whisper for End-to-End Multi-Talker Speech Recognition

链接: https://arxiv.org/abs/2510.03723
作者: Martin Kocour,Martin Karafiat,Alexander Polok,Dominik Klement,Lukáš Burget,Jan Černocký
机构: 未知
类目: Audio and Speech Processing (eess.AS); Computation and Language (cs.CL); Sound (cs.SD)
备注:

点击查看摘要

计算机视觉

[CV-0] Pulp Motion: Framing-aware multimodal camera and human motion generation WWW

链接: https://arxiv.org/abs/2510.05097
作者: Robin Courant,Xi Wang,David Loiseaux,Marc Christie,Vicky Kalogeiton
机构: LIX, École Polytechnique, CNRS, IPP; Inria Saclay; Inria, IRISA, Univ Rennes, CNRS
类目: Graphics (cs.GR); Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL

点击查看摘要

[CV-1] VChain: Chain-of-Visual-Thought for Reasoning in Video Generation

链接: https://arxiv.org/abs/2510.05094
作者: Ziqi Huang,Ning Yu,Gordon Chen,Haonan Qiu,Paul Debevec,Ziwei Liu
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL Code: this https URL

点击查看摘要

[CV-2] Character Mixing for Video Generation

链接: https://arxiv.org/abs/2510.05093
作者: Tingting Liao,Chongjian Ge,Guangyi Liu,Hao Li,Yi Zhou
机构: Mohamed bin Zayed University of Artificial Intelligence (穆罕默德·本·扎耶德人工智能大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-3] Factuality Matters: When Image Generation and Editing Meet Structured Visuals

链接: https://arxiv.org/abs/2510.05091
作者: Le Zhuo,Songhao Han,Yuandong Pu,Boxiang Qiu,Sayak Paul,Yue Liao,Yihao Liu,Jie Shao,Xi Chen,Si Liu,Hongsheng Li
机构: CUHK MMLab(香港中文大学多媒体实验室); Beihang University(北京航空航天大学); Krea AI; Shanghai Jiao Tong University(上海交通大学); Shanghai AI Lab; Hugging Face; National University of Singapore(新加坡国立大学); ByteDance(字节跳动); The University of Hong Kong(香港大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL

点击查看摘要

[CV-4] SAEdit: Token-level control for continuous image editing via Sparse AutoEncoder

【速读】：该论文旨在解决当前大规模文本到图像扩散模型在图像编辑中控制能力不足的问题，尤其是缺乏对特定属性的解耦（disentanglement）和连续可调（continuous control）的能力。解决方案的关键在于通过词元级（token-level）操作文本嵌入（text embeddings），利用稀疏自编码器（Sparse Autoencoder, SAE）识别语义隔离的潜在维度，并沿这些方向对嵌入进行微调，从而实现对目标属性强度的平滑调节。该方法不修改扩散过程，具备模型无关性，适用于多种图像生成架构。

链接: https://arxiv.org/abs/2510.05081
作者: Ronen Kamenetsky,Sara Dorfman,Daniel Garibi,Roni Paiss,Or Patashnik,Daniel Cohen-Or
机构: Tel Aviv University (特拉维夫大学); Google DeepMind (谷歌深度大脑)
类目: Graphics (cs.GR); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: Project page at: this https URL

点击查看摘要

Abstract:Large-scale text-to-image diffusion models have become the backbone of modern image editing, yet text prompts alone do not offer adequate control over the editing process. Two properties are especially desirable: disentanglement, where changing one attribute does not unintentionally alter others, and continuous control, where the strength of an edit can be smoothly adjusted. We introduce a method for disentangled and continuous editing through token-level manipulation of text embeddings. The edits are applied by manipulating the embeddings along carefully chosen directions, which control the strength of the target attribute. To identify such directions, we employ a Sparse Autoencoder (SAE), whose sparse latent space exposes semantically isolated dimensions. Our method operates directly on text embeddings without modifying the diffusion process, making it model agnostic and broadly applicable to various image synthesis backbones. Experiments show that it enables intuitive and efficient manipulations with continuous control across diverse attributes and domains.
zh

[CV-5] Neuroplastic Modular Framework: Cross-Domain Image Classification of Garbage and Industrial Surfaces

【速读】：该论文旨在解决动态环境中图像分类任务中模型泛化能力不足与适应性差的问题，尤其在垃圾分类和工业表面缺陷检测等实际场景中，传统静态模型难以应对数据分布变化和复杂度提升。解决方案的关键在于提出一种神经可塑模块化分类器（Neuroplastic Modular Classifier），其核心创新包括：1）融合ResNet-50与视觉Transformer（Vision Transformer, ViT）的混合架构，兼顾局部特征提取与全局语义建模；2）引入基于FAISS的相似性检索机制，实现对历史样本的记忆式参考以扩展特征空间；3）设计可扩展、可学习的模块化结构，在训练过程中当性能停滞时动态增长，模拟生物学习系统的适应性，从而增强模型对复杂数据的长期适应能力和泛化性能。

链接: https://arxiv.org/abs/2510.05071
作者: Debojyoti Ghosh,Soumya K Ghosh,Adrijit Goswami
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

Abstract:Efficient and accurate classification of waste and industrial surface defects is essential for ensuring sustainable waste management and maintaining high standards in quality control. This paper introduces the Neuroplastic Modular Classifier, a novel hybrid architecture designed for robust and adaptive image classification in dynamic environments. The model combines a ResNet-50 backbone for localized feature extraction with a Vision Transformer (ViT) to capture global semantic context. Additionally, FAISS-based similarity retrieval is incorporated to provide a memory-like reference to previously encountered data, enriching the model’s feature space. A key innovation of our architecture is the neuroplastic modular design composed of expandable, learnable blocks that dynamically grow during training when performance plateaus. Inspired by biological learning systems, this mechanism allows the model to adapt to data complexity over time, improving generalization. Beyond garbage classification, we validate the model on the Kolektor Surface Defect Dataset 2 (KolektorSDD2), which involves industrial defect detection on metal surfaces. Experimental results across domains show that the proposed architecture outperforms traditional static models in both accuracy and adaptability. The Neuroplastic Modular Classifier offers a scalable, high-performance solution for real-world image classification, with strong applicability in both environmental and industrial domains.
zh

[CV-6] StaMo: Unsupervised Learning of Generalizable Robot Motion from Compact State Representation

【速读】：该论文旨在解决具身智能中状态表示（state representation）的表达能力与压缩性难以平衡的问题，即现有方法通常生成冗余或缺乏任务关键信息的状态表征，影响世界建模和决策效率。其解决方案的关键在于提出一种无监督学习框架——StaMo，通过轻量级编码器与预训练扩散 Transformer（Diffusion Transformer, DiT）解码器协同工作，从静态图像中学习高度压缩的双标记状态表示；该表示不仅高效且可解释，还能自然地通过潜在空间插值得到结构化的潜在动作（latent action），无需显式监督即可捕捉系统动力学特性，并可直接映射为机器人执行动作，从而显著提升真实场景任务成功率（提升30%）和政策协同训练性能（提升10.4%）。

链接: https://arxiv.org/abs/2510.05057
作者: Mingyu Liu,Jiuhe Shu,Hui Chen,Zeju Li,Canyu Zhao,Jiange Yang,Shenyuan Gao,Hao Chen,Chunhua Shen
机构: Zhejiang University (浙江大学); Nanjing University (南京大学); Hong Kong University of Science and Technology (香港科技大学)
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

Abstract:A fundamental challenge in embodied intelligence is developing expressive and compact state representations for efficient world modeling and decision making. However, existing methods often fail to achieve this balance, yielding representations that are either overly redundant or lacking in task-critical information. We propose an unsupervised approach that learns a highly compressed two-token state representation using a lightweight encoder and a pre-trained Diffusion Transformer (DiT) decoder, capitalizing on its strong generative prior. Our representation is efficient, interpretable, and integrates seamlessly into existing VLA-based models, improving performance by 14.3% on LIBERO and 30% in real-world task success with minimal inference overhead. More importantly, we find that the difference between these tokens, obtained via latent interpolation, naturally serves as a highly effective latent action, which can be further decoded into executable robot actions. This emergent capability reveals that our representation captures structured dynamics without explicit supervision. We name our method StaMo for its ability to learn generalizable robotic Motion from compact State representation, which is encoded from static images, challenging the prevalent dependence to learning latent action on complex architectures and video data. The resulting latent actions also enhance policy co-training, outperforming prior methods by 10.4% with improved interpretability. Moreover, our approach scales effectively across diverse data sources, including real-world robot data, simulation, and human egocentric video.
zh

[CV-7] No-reference Quality Assessment of Contrast-distorted Images using Contrast-enhanced Pseudo Reference

【速读】：该论文旨在解决对比度失真（contrast distortion）图像质量评估问题，这一问题在现有无参考图像质量评估（No-Reference Image Quality Assessment, NR-IQA）方法中长期被忽视，因其视觉影响机制不同于模糊、噪声等传统失真类型。解决方案的关键在于将NR-IQA问题转化为全参考（Full-Reference, FR）评估：通过一系列对比度增强算法生成伪参考图像（pseudoreference image），使其在视觉上尽可能接近真实参考图像；为此，作者构建了一个大规模对比度增强图像数据集，并训练一个分类网络以根据图像内容和失真特征选择最优的增强算法来生成伪参考图像，最终采用FR方式计算退化图像与伪参考图像之间的质量差异，从而实现更准确的对比度失真图像质量评估。

链接: https://arxiv.org/abs/2510.05053
作者: Mohammad-Ali Mahmoudpour,Saeed Mahmoudpour
机构: Vrije Universiteit Brussel (布鲁塞尔自由大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

Abstract:Contrast change is an important factor that affects the quality of images. During image capturing, unfavorable lighting conditions can cause contrast change and visual quality loss. While various methods have been proposed to assess the quality of images under different distortions such as blur and noise, contrast distortion has been largely overlooked as its visual impact and properties are different from other conventional types of distortions. In this paper, we propose a no-reference image quality assessment (NR-IQA) metric for contrast-distorted images. Using a set of contrast enhancement algorithms, we aim to generate pseudo-reference images that are visually close to the actual reference image, such that the NR problem is transformed to a Full-reference (FR) assessment with higher accuracy. To this end, a large dataset of contrast-enhanced images is produced to train a classification network that can select the most suitable contrast enhancement algorithm based on image content and distortion for pseudo-reference image generation. Finally, the evaluation is performed in the FR manner to assess the quality difference between the contrast-enhanced (pseudoreference) and degraded images. Performance evaluation of the proposed method on three databases containing contrast distortions (CCID2014, TID2013, and CSIQ), indicates the promising performance of the proposed method.
zh

[CV-8] SegMASt3R: Geometry Grounded Segment Matching NEURIPS2025

【速读】：该论文旨在解决宽基线（wide-baseline）图像间的语义或几何一致区域匹配问题，尤其在极端视角变化（最大达180度）下仍保持高精度的段落级匹配（segment matching）。传统关键点匹配方法受限于局部特征的稳定性，难以应对遮挡、光照变化和视角剧烈变动；而本文提出的方法利用3D基础模型（3D foundation models）的空间理解能力，通过其归纳偏置（inductive bias）来增强跨视角段落的语义一致性建模，从而实现更鲁棒的匹配。解决方案的关键在于将3D基础模型的结构先验引入到2D图像段落匹配任务中，显著提升了在ScanNet++和Replica数据集上的AUPRC指标，最高优于现有方法（如SAM2视频传播器和局部特征匹配方法）达30%。

链接: https://arxiv.org/abs/2510.05051
作者: Rohit Jayanti,Swayam Agrawal,Vansh Garg,Siddharth Tourani,Muhammad Haris Khan,Sourav Garg,Madhava Krishna
机构: IIIT Hyderabad (印度信息技术研究所海得拉巴分校); University of Heidelberg (海德堡大学); MBZUAI (穆罕默德·本·扎耶德人工智能大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to The Thirty-Ninth Annual Conference on Neural Information Processing Systems (NeurIPS 2025) as a Spotlight (top 3.5%)

点击查看摘要

Abstract:Segment matching is an important intermediate task in computer vision that establishes correspondences between semantically or geometrically coherent regions across images. Unlike keypoint matching, which focuses on localized features, segment matching captures structured regions, offering greater robustness to occlusions, lighting variations, and viewpoint changes. In this paper, we leverage the spatial understanding of 3D foundation models to tackle wide-baseline segment matching, a challenging setting involving extreme viewpoint shifts. We propose an architecture that uses the inductive bias of these 3D foundation models to match segments across image pairs with up to 180 degree view-point change. Extensive experiments show that our approach outperforms state-of-the-art methods, including the SAM2 video propagator and local feature matching methods, by upto 30% on the AUPRC metric, on ScanNet++ and Replica datasets. We further demonstrate benefits of the proposed model on relevant downstream tasks, including 3D instance segmentation and image-goal navigation. Project Page: this https URL
zh

[CV-9] Video-LMM Post-Training: A Deep Dive into Video Reasoning with Large Multimodal Models

【速读】：该论文旨在解决视频理解领域中生成式 AI（Generative AI）模型在后训练（post-training）阶段方法碎片化的问题，特别是如何将基础感知模型转化为具备复杂推理能力的系统。其核心解决方案在于提出一个结构化的分类体系，涵盖三个关键支柱：基于链式思维（chain-of-thought）的监督微调（SFT）、基于可验证目标的强化学习（Reinforcement Learning, RL），以及通过增强推理计算实现的测试时扩展（Test-Time Scaling, TTS）。该框架不仅厘清了各类方法的角色与相互关系，还针对视频特有的挑战（如时间定位、时空定位、长视频效率及多模态证据融合）提出了适配策略，并系统梳理了评估协议与基准数据集，为提升视频大模型（Video-Large Multimodal Models, Video-LMMs）的推理能力提供了统一的研究路径。

链接: https://arxiv.org/abs/2510.05034
作者: Yunlong Tang,Jing Bi,Pinxin Liu,Zhenyu Pan,Zhangyun Tan,Qianxiang Shen,Jiani Liu,Hang Hua,Junjia Guo,Yunzhong Xiao,Chao Huang,Zhiyuan Wang,Susan Liang,Xinyi Liu,Yizhi Song,Yuhe Nie,Jia-Xing Zhong,Bozheng Li,Daiqing Qi,Ziyun Zeng,Ali Vosoughi,Luchuan Song,Zeliang Zhang,Daiki Shimada,Han Liu,Jiebo Luo,Chenliang Xu
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: The 1st version

点击查看摘要

Abstract:Video understanding represents the most challenging frontier in computer vision, requiring models to reason about complex spatiotemporal relationships, long-term dependencies, and multimodal evidence. The recent emergence of Video-Large Multimodal Models (Video-LMMs), which integrate visual encoders with powerful decoder-based language models, has demonstrated remarkable capabilities in video understanding tasks. However, the critical phase that transforms these models from basic perception systems into sophisticated reasoning engines, post-training, remains fragmented across the literature. This survey provides the first comprehensive examination of post-training methodologies for Video-LMMs, encompassing three fundamental pillars: supervised fine-tuning (SFT) with chain-of-thought, reinforcement learning (RL) from verifiable objectives, and test-time scaling (TTS) through enhanced inference computation. We present a structured taxonomy that clarifies the roles, interconnections, and video-specific adaptations of these techniques, addressing unique challenges such as temporal localization, spatiotemporal grounding, long video efficiency, and multimodal evidence integration. Through systematic analysis of representative methods, we synthesize key design principles, insights, and evaluation protocols while identifying critical open challenges in reward design, scalability, and cost-performance optimization. We further curate essential benchmarks, datasets, and metrics to facilitate rigorous assessment of post-training effectiveness. This survey aims to provide researchers and practitioners with a unified framework for advancing Video-LMM capabilities. Additional resources and updates are maintained at: this https URL
zh

[CV-10] Exploring the Efficacy of Modified Transfer Learning in Identifying Parkinsons Disease Through Drawn Image Patterns

【速读】：该论文旨在解决帕金森病（Parkinson’s disease, PD）早期诊断困难的问题，传统方法存在操作繁琐和成本高等局限性。其解决方案的关键在于利用手绘螺旋图和波浪图作为潜在生物标志物，结合卷积神经网络（convolutional neural networks, CNNs）、迁移学习与注意力机制提升模型性能并增强抗过拟合能力；同时通过数据增强扩充训练集多样性，并采用三阶段架构（预训练CNN、自定义卷积层及集成投票）实现高精度分类，最终在螺旋图和波浪图上分别达到90%和96.67%的加权平均F1分数，集成硬投票后整体准确率达93.3%，展现出机器学习在PD早期非侵入式、低成本诊断中的显著潜力。

链接: https://arxiv.org/abs/2510.05015
作者: Nabil Daiyan,Md Rakibul Haque
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 11 figures, published on 2024 2nd International Conference on Information and Communication Technology (ICICT 2024)

点击查看摘要

Abstract:Parkinson’s disease (PD) is a progressive neurodegenerative condition characterized by the death of dopaminergic neurons, leading to various movement disorder symptoms. Early diagnosis of PD is crucial to prevent adverse effects, yet traditional diagnostic methods are often cumbersome and costly. In this study, a machine learning-based approach is proposed using hand-drawn spiral and wave images as potential biomarkers for PD detection. Our methodology leverages convolutional neural networks (CNNs), transfer learning, and attention mechanisms to improve model performance and resilience against overfitting. To enhance the diversity and richness of both spiral and wave categories, the training dataset undergoes augmentation to increase the number of images. The proposed architecture comprises three phases: utilizing pre-trained CNNs, incorporating custom convolutional layers, and ensemble voting. Employing hard voting further enhances performance by aggregating predictions from multiple models. Experimental results show promising accuracy rates. For spiral images, weighted average precision, recall, and F1-score are 90%, and for wave images, they are 96.67%. After combining the predictions through ensemble hard voting, the overall accuracy is 93.3%. These findings underscore the potential of machine learning in early PD diagnosis, offering a non-invasive and cost-effective solution to improve patient outcomes.
zh

[CV-11] Latent Uncertainty Representations for Video-based Driver Action and Intention Recognition

【速读】：该论文旨在解决资源受限环境中深度神经网络（Deep Neural Networks, DNNs）在安全关键任务（如基于视频的驾驶员行为与意图识别）中对分布外（Out-of-Distribution, OOD）样本检测能力不足的问题。现有最后一层概率深度学习（Last Layer Probabilistic Deep Learning, LL-PDL）方法虽能实现OOD检测，但性能不稳定。其解决方案的关键在于：通过在预训练DNN基础上增加变换层（transformation layers），生成多个潜在表示（latent representations）以估计不确定性，提出潜不确定性表示（Latent Uncertainty Representation, LUR）及其改进版本——排斥训练LUR（Repulsively Trained LUR, RLUR）。该方法在保持与主流LL-PDL方法相当的分布内分类性能的同时，在OOD检测上达到顶尖水平，且训练效率更高、调参更简便，无需依赖马尔可夫链蒙特卡洛（Markov-Chain Monte Carlo, MCMC）采样或复杂的排斥训练过程。

链接: https://arxiv.org/abs/2510.05006
作者: Koen Vellenga,H. Joe Steinhauer,Jonas Andersson,Anders Sjögren
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 16 pages, 8 figures, 7 tables, under submission

点击查看摘要

Abstract:Deep neural networks (DNNs) are increasingly applied to safety-critical tasks in resource-constrained environments, such as video-based driver action and intention recognition. While last layer probabilistic deep learning (LL-PDL) methods can detect out-of-distribution (OOD) instances, their performance varies. As an alternative to last layer approaches, we propose extending pre-trained DNNs with transformation layers to produce multiple latent representations to estimate the uncertainty. We evaluate our latent uncertainty representation (LUR) and repulsively trained LUR (RLUR) approaches against eight PDL methods across four video-based driver action and intention recognition datasets, comparing classification performance, calibration, and uncertainty-based OOD detection. We also contribute 28,000 frame-level action labels and 1,194 video-level intention labels for the NuScenes dataset. Our results show that LUR and RLUR achieve comparable in-distribution classification performance to other LL-PDL approaches. For uncertainty-based OOD detection, LUR matches top-performing PDL methods while being more efficient to train and easier to tune than approaches that require Markov-Chain Monte Carlo sampling or repulsive training procedures.
zh

[CV-12] Bridging Text and Video Generation: A Survey

链接: https://arxiv.org/abs/2510.04999
作者: Nilay Kumar,Priyansh Bhandari,G. Maragatham
机构: SRM Institute of Science and Technology, KTR (SRM科技大学，KTR校区)
类目: Graphics (cs.GR); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-13] ActiveMark: on watermarking of visual foundation models via massive activations

【速读】：该论文旨在解决视觉基础模型（Visual Foundation Models, VFMs）在知识产权保护中的所有权验证问题，即如何有效区分被授权使用的模型与非法复制并重新分发的模型。解决方案的关键在于通过微调VFMs中少量具有表达力的层，并结合一个小型编码器-解码器网络，在保留模型功能性的前提下，将数字水印嵌入到一组预留输入图像的内部表示中。该方法确保水印在模型经过下游任务微调后依然可检测，从而实现对模型合法性的可靠验证，且理论和实验均表明其具备低误检率和低漏检率。

链接: https://arxiv.org/abs/2510.04966
作者: Anna Chistyakova,Mikhail Pautov
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

Abstract:Being trained on large and vast datasets, visual foundation models (VFMs) can be fine-tuned for diverse downstream tasks, achieving remarkable performance and efficiency in various computer vision applications. The high computation cost of data collection and training motivates the owners of some VFMs to distribute them alongside the license to protect their intellectual property rights. However, a dishonest user of the protected model’s copy may illegally redistribute it, for example, to make a profit. As a consequence, the development of reliable ownership verification tools is of great importance today, since such methods can be used to differentiate between a redistributed copy of the protected model and an independent model. In this paper, we propose an approach to ownership verification of visual foundation models by fine-tuning a small set of expressive layers of a VFM along with a small encoder-decoder network to embed digital watermarks into an internal representation of a hold-out set of input images. Importantly, the watermarks embedded remain detectable in the functional copies of the protected model, obtained, for example, by fine-tuning the VFM for a particular downstream task. Theoretically and experimentally, we demonstrate that the proposed method yields a low probability of false detection of a non-watermarked model and a low probability of false misdetection of a watermarked model.
zh

[CV-14] SSDD: Single-Step Diffusion Decoder for Efficient Image Tokenization

链接: https://arxiv.org/abs/2510.04961
作者: Théophane Vallaeys,Jakob Verbeek,Matthieu Cord
机构: Meta(Meta)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-15] Bidirectional Mammogram View Translation with Column-Aware and Implicit 3D Conditional Diffusion

链接: https://arxiv.org/abs/2510.04947
作者: Xin Li,Kaixiang Yang,Qiang Li,Zhiwei Wang
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: BIBM2025 accept, 8 pages, 4 figures

点击查看摘要

[CV-16] Unsupervised Active Learning via Natural Feature Progressive Framework

【速读】：该论文旨在解决无监督主动学习（Unsupervised Active Learning, UAL）中样本选择效率与性能不足的问题。现有UAL方法通常依赖局部梯度评分来估计样本重要性，易受噪声和模糊数据干扰，且难以覆盖完整数据分布，同时采用单次线性筛选策略，无法实现真正的无监督主动学习范式。其解决方案的关键在于提出自然特征渐进框架（Natural Feature Progressive Framework, NFPF），该框架引入特定特征学习机（Specific Feature Learning Machine, SFLM）以精准量化每个样本对模型性能的贡献，并基于SFLM构建强大的重构差异指标用于初始样本选择，从而显著提升样本代表性、鲁棒性和整体性能，最终在视觉数据集上达到与监督式主动学习相当的效果。

链接: https://arxiv.org/abs/2510.04939
作者: Yuxi Liu,Catherine Lalman,Yimin Yang
机构: Western University (西门大学); Thomas Jefferson University (托马斯·杰斐逊大学); Vector Institute (向量研究所)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Under review at IEEE TPAMI

点击查看摘要

Abstract:The effectiveness of modern deep learning models is predicated on the availability of large-scale, human-annotated datasets, a process that is notoriously expensive and time-consuming. While Active Learning (AL) offers a strategic solution by labeling only the most informative and representative data, its iterative nature still necessitates significant human involvement. Unsupervised Active Learning (UAL) presents an alternative by shifting the annotation burden to a single, post-selection step. Unfortunately, prevailing UAL methods struggle to achieve state-of-the-art performance. These approaches typically rely on local, gradient-based scoring for sample importance estimation, which not only makes them vulnerable to ambiguous and noisy data but also hinders their capacity to select samples that adequately represent the full data distribution. Moreover, their use of shallow, one-shot linear selection falls short of a true UAL paradigm. In this paper, we propose the Natural Feature Progressive Framework (NFPF), a UAL method that revolutionizes how sample importance is measured. At its core, NFPF employs a Specific Feature Learning Machine (SFLM) to effectively quantify each sample’s contribution to model performance. We further utilize the SFLM to define a powerful Reconstruction Difference metric for initial sample selection. Our comprehensive experiments show that NFPF significantly outperforms all established UAL methods and achieves performance on par with supervised AL methods on vision datasets. Detailed ablation studies and qualitative visualizations provide compelling evidence for NFPF’s superior performance, enhanced robustness, and improved data distribution coverage.
zh

[CV-17] REN: Anatomically-Informed Mixture-of-Experts for Interstitial Lung Disease Diagnosis

链接: https://arxiv.org/abs/2510.04923
作者: Alec K. Peltekian,Halil Ertugrul Aktas,Gorkem Durak,Kevin Grudzinski,Bradford C. Bemiss,Carrie Richardson,Jane E. Dematte,G. R. Scott Budinger,Anthony J. Esposito,Alexander Misharin,Alok Choudhary,Ankit Agrawal,Ulas Bagci
机构: Northwestern University McCormick School of Engineering and Applied Science (西北大学麦考密克工程与应用科学学院); Northwestern University Feinberg School of Medicine (西北大学费恩伯格医学院); Simpson Querrey Lung Institute for Translational Science (辛普森-奎雷肺部转化科学研究所)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 10 pages, 4 figures, 2 tables

点击查看摘要

[CV-18] A Semantics-Aware Hierarchical Self-Supervised Approach to Classification of Remote Sensing Images

【速读】：该论文旨在解决遥感图像分类中忽略类别层次结构（hierarchical structure）的问题，传统方法多聚焦于细粒度分类而未充分利用类间语义关系。其解决方案的关键在于提出一种语义感知的层次共识（Semantics-Aware Hierarchical Consensus, SAHC）方法：通过在深度网络架构中集成针对不同粒度层级的分类头（hierarchy-specific classification heads），并引入可训练的层次矩阵（hierarchy matrices）以自监督方式引导网络学习层次结构；同时设计层次共识机制（hierarchical consensus mechanism），作为加权集成策略确保不同层级概率分布的一致性，从而有效利用层次分类任务的内在结构，提升模型在多种复杂度下的适应性和鲁棒性。

链接: https://arxiv.org/abs/2510.04916
作者: Giulio Weikmann,Gianmarco Perantoni,Lorenzo Bruzzone
机构: University of Trento (特伦托大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 12 pages, 6 figures

点击查看摘要

Abstract:Deep learning has become increasingly important in remote sensing image classification due to its ability to extract semantic information from complex data. Classification tasks often include predefined label hierarchies that represent the semantic relationships among classes. However, these hierarchies are frequently overlooked, and most approaches focus only on fine-grained classification schemes. In this paper, we present a novel Semantics-Aware Hierarchical Consensus (SAHC) method for learning hierarchical features and relationships by integrating hierarchy-specific classification heads within a deep network architecture, each specialized in different degrees of class granularity. The proposed approach employs trainable hierarchy matrices, which guide the network through the learning of the hierarchical structure in a self-supervised manner. Furthermore, we introduce a hierarchical consensus mechanism to ensure consistent probability distributions across different hierarchical levels. This mechanism acts as a weighted ensemble being able to effectively leverage the inherent structure of the hierarchical classification task. The proposed SAHC method is evaluated on three benchmark datasets with different degrees of hierarchical complexity on different tasks, using distinct backbone architectures to effectively emphasize its adaptability. Experimental results show both the effectiveness of the proposed approach in guiding network learning and the robustness of the hierarchical consensus for remote sensing image classification tasks.
zh

[CV-19] Comparative Analysis of YOLOv5 Faster R-CNN SSD and RetinaNet for Motorbike Detection in Kigali Autonomous Driving Context

链接: https://arxiv.org/abs/2510.04912
作者: Ngeyen Yinkfu,Sunday Nwovu,Jonathan Kayizzi,Angelique Uwamahoro
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 3 figures, 2 tables

点击查看摘要

[CV-20] CLEAR-IR: Clarity-Enhanced Active Reconstruction of Infrared Imagery

链接: https://arxiv.org/abs/2510.04883
作者: Nathan Shankar,Pawel Ladosz,Hujun Yin
机构: 未知
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 8 pages, 8 figures

点击查看摘要

[CV-21] BenthiCat: An opti-acoustic dataset for advancing benthic classification and habitat mapping

【速读】：该论文旨在解决海底生境制图领域中因高质量标注数据稀缺而导致的机器学习模型发展与基准测试受限的问题。其解决方案的关键在于构建了一个大规模、多模态的数据集，包含约一百万个侧扫声呐（Side-Scan Sonar, SSS）图像块，辅以地形高程图和由自主水下航行器（Autonomous Underwater Vehicle, AUV）获取的共注册光学影像，并对其中约36,000个SSS图像块进行人工分割标注，以支持监督微调分类模型。此外，通过空间关联光学图像与对应SSS图像块，实现了自监督跨模态表示学习，从而促进多传感器融合算法的发展，为水下生境自动分类与多源数据集成提供标准化基准资源。

链接: https://arxiv.org/abs/2510.04876
作者: Hayat Rajani,Valerio Franchi,Borja Martinez-Clavel Valles,Raimon Ramos,Rafael Garcia,Nuno Gracias
机构: University of Girona (赫罗纳大学); Tecnoambiente SLU
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Article under review by IJRR

点击查看摘要

Abstract:Benthic habitat mapping is fundamental for understanding marine ecosystems, guiding conservation efforts, and supporting sustainable resource management. Yet, the scarcity of large, annotated datasets limits the development and benchmarking of machine learning models in this domain. This paper introduces a thorough multi-modal dataset, comprising about a million side-scan sonar (SSS) tiles collected along the coast of Catalonia (Spain), complemented by bathymetric maps and a set of co-registered optical images from targeted surveys using an autonomous underwater vehicle (AUV). Approximately \num36000 of the SSS tiles have been manually annotated with segmentation masks to enable supervised fine-tuning of classification models. All the raw sensor data, together with mosaics, are also released to support further exploration and algorithm development. To address challenges in multi-sensor data fusion for AUVs, we spatially associate optical images with corresponding SSS tiles, facilitating self-supervised, cross-modal representation learning. Accompanying open-source preprocessing and annotation tools are provided to enhance accessibility and encourage research. This resource aims to establish a standardized benchmark for underwater habitat mapping, promoting advancements in autonomous seafloor classification and multi-sensor integration.
zh

[CV-22] In-Field Mapping of Grape Yield and Quality with Illumination-Invariant Deep Learning

链接: https://arxiv.org/abs/2510.04864
作者: Ciem Cornelissen,Sander De Coninck,Axel Willekens,Sam Leroux,Pieter Simoens
机构: IDLab, Department of Information Technology at Ghent University - imec (IDLab，根特大学信息科技系 - imec); Ghent University, IDLab ‐ Imec ‐ AI and Robotics Lab (AIRO) (根特大学，IDLab ‐ imec ‐ 人工智能与机器人实验室 (AIRO)); Institute for Agriculture, Food and Fishery Flanders (ILVO) (弗拉芒农业、食品与渔业研究所 (ILVO))
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted manuscript for the IEEE Internet of Things Journal. The final version will be available on IEEE Xplore. \c{opyright} 2025 IEEE

点击查看摘要

[CV-23] μDeepIQA: deep learning-based fast and robust image quality assessment with local predictions for optical microscopy

链接: https://arxiv.org/abs/2510.04859
作者: Elena Corbetta,Thomas Bocklitz
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Data Analysis, Statistics and Probability (physics.data-an); Quantitative Methods (q-bio.QM)
备注: 16 pages, 6 figures. μDeepIQA is publicly available at this https URL

点击查看摘要

[CV-24] ERDE: Entropy-Regularized Distillation for Early-exit

链接: https://arxiv.org/abs/2510.04856
作者: Martial Guidez,Stefan Duffner,Yannick Alpou,Oscar Röth,Christophe Garcia
机构: INSA Lyon (法国国家高等工程技术学院); CNRS (法国国家科学研究中心); Université Claude Bernard Lyon 1 (克莱蒙-奥弗涅大学); LIRIS, UMR5205 (信息、计算机与应用数学实验室)
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:

点击查看摘要

[CV-25] Read the Room: Inferring Social Context Through Dyadic Interaction Recognition in Cyber-physical-social Infrastructure Systems

【速读】：该论文旨在解决当前网络物理系统（Cyber-Physical Systems, CPS）在追求经济目标（如性能与安全）时忽视人类中心（或“社会”）效益的问题，提出构建网络物理社会基础设施系统（Cyber-Physical-Social Infrastructure Systems, CPSIS）以实现社会目标对齐。其解决方案的关键在于通过真实世界数据识别二人互动（dyadic human interactions），作为测量社会行为的基础，并采用深度传感器替代RGB摄像头以保障隐私，同时利用五种基于骨骼的动作识别算法对12类典型二人互动（如象征性手势和情感表达）进行比较分析，从而揭示人际互动中的文化与情感内涵，为预测和引导积极社会结果提供建模依据。

链接: https://arxiv.org/abs/2510.04854
作者: Cheyu Lin,John Martins,Katherine A. Flanigan,Ph.D
机构: Carnegie Mellon University (卡内基梅隆大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: ASCE International Conference on Computing in Civil Engineering 2024

点击查看摘要

Abstract:Cyber-physical systems (CPS) integrate sensing, computing, and control to improve infrastructure performance, focusing on economic goals like performance and safety. However, they often neglect potential human-centered (or ‘‘social’’) benefits. Cyber-physical-social infrastructure systems (CPSIS) aim to address this by aligning CPS with social objectives. This involves defining social benefits, understanding human interactions with each other and infrastructure, developing privacy-preserving measurement methods, modeling these interactions for prediction, linking them to social benefits, and actuating the physical environment to foster positive social outcomes. This paper delves into recognizing dyadic human interactions using real-world data, which is the backbone to measuring social behavior. This lays a foundation to address the need to enhance understanding of the deeper meanings and mutual responses inherent in human interactions. While RGB cameras are informative for interaction recognition, privacy concerns arise. Depth sensors offer a privacy-conscious alternative by analyzing skeletal movements. This study compares five skeleton-based interaction recognition algorithms on a dataset of 12 dyadic interactions. Unlike single-person datasets, these interactions, categorized into communication types like emblems and affect displays, offer insights into the cultural and emotional aspects of human interactions.
zh

[CV-26] From Actions to Kinesics: Extracting Human Psychological States through Bodily Movements ALT

【速读】：该论文旨在解决人类与建成环境之间动态交互关系建模中的核心难题，即如何在保证泛化能力和隐私安全的前提下，准确捕捉人类的心理状态。传统方法依赖理论模型或问卷调查，存在范围有限、静态且劳动密集的缺陷。解决方案的关键在于提出一种基于肢体动作识别（kinesics）的框架，通过结合时空图卷积网络（ST-GCN）与卷积神经网络（CNN），利用迁移学习从3D骨骼关节点数据中直接推断行为的交际功能，从而无需人工定义物理动作与心理类别的映射关系，既保留了用户匿名性，又揭示了反映认知和情绪状态的潜在运动结构。

链接: https://arxiv.org/abs/2510.04844
作者: Cheyu Lin,Katherine A. Flanigan
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: The 15th International Workshop on Structural Health Monitoring (IWSHM)

点击查看摘要

Abstract:Understanding the dynamic relationship between humans and the built environment is a key challenge in disciplines ranging from environmental psychology to reinforcement learning (RL). A central obstacle in modeling these interactions is the inability to capture human psychological states in a way that is both generalizable and privacy preserving. Traditional methods rely on theoretical models or questionnaires, which are limited in scope, static, and labor intensive. We present a kinesics recognition framework that infers the communicative functions of human activity – known as kinesics – directly from 3D skeleton joint data. Combining a spatial-temporal graph convolutional network (ST-GCN) with a convolutional neural network (CNN), the framework leverages transfer learning to bypass the need for manually defined mappings between physical actions and psychological categories. The approach preserves user anonymity while uncovering latent structures in bodily movements that reflect cognitive and emotional states. Our results on the Dyadic User EngagemenT (DUET) dataset demonstrate that this method enables scalable, accurate, and human-centered modeling of behavior, offering a new pathway for enhancing RL-driven simulations of human-environment interaction.
zh

[CV-27] Detailed Aerial Mapping of Photovoltaic Power Plants Through Semantically Significant Keypoints

链接: https://arxiv.org/abs/2510.04840
作者: Viktor Kozák,Jan Chudoba,Libor Přeučil
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 18 figures

点击查看摘要

[CV-28] Beyond Random: Automatic Inner-loop Optimization in Dataset Distillation

链接: https://arxiv.org/abs/2510.04838
作者: Muquan Li,Hang Gou,Dongyang Zhang,Shuang Liang,Xiurui Xie,Deqiang Ouyang,Ke Qin
机构: University of Electronic Science and Technology of China (电子科技大学); Chongqing University (重庆大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:

点击查看摘要

[CV-29] Flow Matching for Conditional MRI-CT and CBCT-CT Image Synthesis

链接: https://arxiv.org/abs/2510.04823
作者: Arnela Hadzic,Simon Johannes Joham,Martin Urschler
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-30] AvatarVTON: 4D Virtual Try-On for Animatable Avatars

链接: https://arxiv.org/abs/2510.04822
作者: Zicheng Jiang,Jixin Gao,Shengfeng He,Xinzhe Li,Yulong Zheng,Zhaotong Yang,Junyu Dong,Yong Du
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-31] Did you just see that? Arbitrary view synthesis for egocentric replay of operating room workflows from ambient sensors

链接: https://arxiv.org/abs/2510.04802
作者: Han Zhang,Lalithkumar Seenivasan,Jose L. Porras,Roger D. Soberanis-Mukul,Hao Ding,Hongchao Shu,Benjamin D. Killeen,Ankita Ghosh,Lonny Yarmus,Masaru Ishii,Angela Christine Argento,Mathias Unberath
机构: Johns Hopkins University (约翰霍普金斯大学); Johns Hopkins Medicine (约翰霍普金斯医学院)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[CV-32] DiT-VTON: Diffusion Transformer Framework for Unified Multi-Category Virtual Try-On and Virtual Try-All with Integrated Image Editing CVPR2025

链接: https://arxiv.org/abs/2510.04797
作者: Qi Li,Shuwen Qiu,Julien Han,Xingzi Xu,Mehmet Saygin Seyfioglu,Kee Kiat Koo,Karim Bouyarmane
机构: Amazon; University of California, Los Angeles (UCLA); Duke University
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Submitted to CVPR 2025 and Published at CVPR 2025 AI for Content Creation workshop

点击查看摘要

[CV-33] A Comparative Study of Vision Transformers and CNNs for Few-Shot Rigid Transformation and Fundamental Matrix Estimation

链接: https://arxiv.org/abs/2510.04794
作者: Alon Kaya,Igal Bilik,Inna Stainvas
机构: Ben-Gurion University of the Negev (本古里安大学); GE Healthcare Science and Technology (GE医疗科学与技术)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-34] Hands-Free Heritage: Automated 3D Scanning for Cultural Heritage Digitization

链接: https://arxiv.org/abs/2510.04781
作者: Javed Ahmad,Federico Dassiè,Selene Frascella,Gabriele Marchello,Ferdinando Cannella,Arianna Traviglia
机构: Centre for Cultural Heritage Technology, Italian Institute of Technology (意大利技术研究院文化遗产技术中心); Industrial Robotics Facility, Italian Institute of Technology (意大利技术研究院工业机器人设施)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 9 pages

点击查看摘要

[CV-35] Federated Learning for Surgical Vision in Appendicitis Classification: Results of the FedSurg EndoVis 2024 Challenge

【速读】：该论文旨在解决联邦学习（Federated Learning, FL）在手术视频分类任务中的泛化能力与本地适应性问题，特别是在多中心临床数据场景下如何实现模型性能提升而不共享患者隐私数据。其关键解决方案在于设计并实施了FedSurg挑战，评估不同FL策略在两个核心任务上的表现：一是跨中心的泛化能力（即模型在未见临床中心的表现），二是通过本地微调后的个性化适应能力。参赛方案涵盖基于ViViT的预训练模型线性探测、度量学习结合三元组损失（triplet loss）以及多种聚合机制（如FedAvg、FedMedian、FedSAM），最终发现架构选择（如ViViT）、时空建模和上下文感知预处理对提升性能至关重要，同时揭示了类不平衡敏感性和超参数调优难度等现实挑战，为未来开发更具鲁棒性、自适应性和抗不平衡性的联邦学习方法提供了基准参考。

链接: https://arxiv.org/abs/2510.04772
作者: Max Kirchner,Hanna Hoffmann,Alexander C. Jenke,Oliver L. Saldanha,Kevin Pfeiffer,Weam Kanjo,Julia Alekseenko,Claas de Boer,Santhi Raj Kolamuri,Lorenzo Mazza,Nicolas Padoy,Sophia Bano,Annika Reinke,Lena Maier-Hein,Danail Stoyanov,Jakob N. Kather,Fiona R. Kolbinger,Sebastian Bodenstedt,Stefanie Speidel
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: A challenge report pre-print (31 pages), including 7 tables and 8 figures

点击查看摘要

Abstract:Purpose: The FedSurg challenge was designed to benchmark the state of the art in federated learning for surgical video classification. Its goal was to assess how well current methods generalize to unseen clinical centers and adapt through local fine-tuning while enabling collaborative model development without sharing patient data. Methods: Participants developed strategies to classify inflammation stages in appendicitis using a preliminary version of the multi-center Appendix300 video dataset. The challenge evaluated two tasks: generalization to an unseen center and center-specific adaptation after fine-tuning. Submitted approaches included foundation models with linear probing, metric learning with triplet loss, and various FL aggregation schemes (FedAvg, FedMedian, FedSAM). Performance was assessed using F1-score and Expected Cost, with ranking robustness evaluated via bootstrapping and statistical testing. Results: In the generalization task, performance across centers was limited. In the adaptation task, all teams improved after fine-tuning, though ranking stability was low. The ViViT-based submission achieved the strongest overall performance. The challenge highlighted limitations in generalization, sensitivity to class imbalance, and difficulties in hyperparameter tuning in decentralized training, while spatiotemporal modeling and context-aware preprocessing emerged as promising strategies. Conclusion: The FedSurg Challenge establishes the first benchmark for evaluating FL strategies in surgical video classification. Findings highlight the trade-off between local personalization and global robustness, and underscore the importance of architecture choice, preprocessing, and loss design. This benchmarking offers a reference point for future development of imbalance-aware, adaptive, and robust FL methods in clinical surgical AI.
zh

[CV-36] Beyond the Seen: Bounded Distribution Estimation for Open-Vocabulary Learning

链接: https://arxiv.org/abs/2510.04770
作者: Xiaomeng Fan,Yuchuan Mao,Zhi Gao,Yuwei Wu,Jin Chen,Yunde Jia
机构: Beijing Key Laboratory of Intelligent Information Technology, School of Computer Science & Technology, Beijing Institute of Technology; Guangdong Laboratory of Machine Perception and Intelligent Computing, Shenzhen MSU-BIT University
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:

点击查看摘要

[CV-37] Progressive Gaussian Transformer with Anisotropy-aware Sampling for Open Vocabulary Occupancy Prediction

链接: https://arxiv.org/abs/2510.04759
作者: Chi Yan,Dan Xu
机构: The Hong Kong University of Science and Technology (香港科技大学); ZEEKR Automobile R&D Co., Ltd (极氪汽车研发有限公司)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Project Page: this https URL

点击查看摘要

[CV-38] Beyond Appearance: Transformer-based Person Identification from Conversational Dynamics

链接: https://arxiv.org/abs/2510.04753
作者: Masoumeh Chapariniya,Teodora Vukovic,Sarah Ebling,Volker Dellwo
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-39] Anomaly-Aware YOLO: A Frugal yet Robust Approach to Infrared Small Target Detection

【速读】：该论文旨在解决红外小目标检测（Infrared Small Target Detection, IRSTD）中因复杂背景和目标尺寸微小导致传统目标检测器误报率高的问题。解决方案的关键在于提出一种异常感知的YOLO模型（Anomaly-Aware YOLO, AA-YOLO），其核心创新是在检测头中引入统计异常检测测试，将小目标视为相对于背景的异常模式，从而有效控制误报率。该方法仅修改检测头，保持骨干网络不变，具有良好的通用性，并在数据有限、噪声干扰及域偏移等实际场景下表现出优异鲁棒性。

链接: https://arxiv.org/abs/2510.04741
作者: Alina Ciocarlan,Sylvie Le Hégarat-Mascle,Sidonie Lefebvre
机构: École Polytechnique (巴黎综合理工学院)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

Abstract:Infrared Small Target Detection (IRSTD) is a challenging task in defense applications, where complex backgrounds and tiny target sizes often result in numerous false alarms using conventional object detectors. To overcome this limitation, we propose Anomaly-Aware YOLO (AA-YOLO), which integrates a statistical anomaly detection test into its detection head. By treating small targets as unexpected patterns against the background, AA-YOLO effectively controls the false alarm rate. Our approach not only achieves competitive performance on several IRSTD benchmarks, but also demonstrates remarkable robustness in scenarios with limited training data, noise, and domain shifts. Furthermore, since only the detection head is modified, our design is highly generic and has been successfully applied across various YOLO backbones, including lightweight models. It also provides promising results when integrated into an instance segmentation YOLO. This versatility makes AA-YOLO an attractive solution for real-world deployments where resources are constrained. The code will be publicly released.
zh

[CV-40] ExposureEngine: Oriented Logo Detection and Sponsor Visibility Analytics in Sports Broadcasts

链接: https://arxiv.org/abs/2510.04739
作者: Mehdi Houshmand Sarkhoosh,Frøy Øye,Henrik Nestor Sørlie,Nam Hoang Vu,Dag Johansen,Cise Midoglu,Tomas Kupka,Pål Halvorsen
机构: 34,134,3,5,4,3,4,134
类目: Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
备注: This work has been submitted to the IEEE for possible publication

点击查看摘要

[CV-41] Benchmark on Monocular Metric Depth Estimation in Wildlife Setting

链接: https://arxiv.org/abs/2510.04723
作者: Niccolò Niccoli,Lorenzo Seidenari,Ilaria Greco,Francesco Rovero
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-42] Object-Centric Representation Learning for Enhanced 3D Scene Graph Prediction NEURIPS2025

链接: https://arxiv.org/abs/2510.04714
作者: KunHo Heo,GiHyun Kim,SuYeon Kim,MyeongAh Cho
机构: Kyung Hee University (中央大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by NeurIPS 2025. Code: this https URL

点击查看摘要

[CV-43] ReactDiff: Fundamental Multiple Appropriate Facial Reaction Diffusion Model

链接: https://arxiv.org/abs/2510.04712
作者: Luo Cheng,Song Siyang,Yan Siyuan,Yu Zhen,Ge Zongyuan
机构: Monash University(蒙纳士大学); University of Exeter(埃克塞特大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC); Multimedia (cs.MM)
备注: Accepted to ACM Multimedia

点击查看摘要

[CV-44] ID-Consistent Precise Expression Generation with Blendshape-Guided Diffusion ICCV

链接: https://arxiv.org/abs/2510.04706
作者: Foivos Paraperas Papantoniou,Stefanos Zafeiriou
机构: Imperial College London (帝国理工学院)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: ICCVW 2025, Code: this https URL

点击查看摘要

[CV-45] Label-Efficient Cross-Modality Generalization for Liver Segmentation in Multi-Phase MRI

链接: https://arxiv.org/abs/2510.04705
作者: Quang-Khai Bui-Tran,Minh-Toan Dinh,Thanh-Huy Nguyen,Ba-Thinh Lam,Mai-Anh Vu,Ulas Bagci
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 11 pages, 3 figures

点击查看摘要

[CV-46] Watch and Learn: Learning to Use Computers from Online Videos

链接: https://arxiv.org/abs/2510.04673
作者: Chan Hee Song,Yiwen Song,Palash Goyal,Yu Su,Oriana Riva,Hamid Palangi,Tomas Pfister
机构: Google Cloud AI Research(谷歌云人工智能研究); The Ohio State University(俄亥俄州立大学); Tomas Pfister
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-47] ConceptSplit: Decoupled Multi-Concept Personalization of Diffusion Models via Token-wise Adaptation and Attention Disentanglement ICCV2025

链接: https://arxiv.org/abs/2510.04668
作者: Habin Lim,Yeongseob Won,Juwon Seo,Gyeong-Moon Park
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 14 pages, 13 figures, to be published in ICCV 2025

点击查看摘要

[CV-48] MoME: Estimating Psychological Traits from Gait with Multi-Stage Mixture of Movement Experts

链接: https://arxiv.org/abs/2510.04654
作者: Andy Cǎtrunǎ,Adrian Cosma,Emilian Rǎdoi
机构: University Politehnica of Bucharest (布加勒斯特理工大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 4 Figures, 4 Tables

点击查看摘要

[CV-49] EduPersona: Benchmarking Subjective Ability Boundaries of Virtual Student Agents

链接: https://arxiv.org/abs/2510.04648
作者: Buyuan Zhu,Shiyu Hu,Yiping Ma,Yuanming Zhang,Kang Hao Cheong
机构: Nanyang Technological University (南洋理工大学); East China Normal University (华东师范大学); Harbin Institute of Technology (哈尔滨工业大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computers and Society (cs.CY)
备注: Preprint, Under review

点击查看摘要

[CV-50] Do Superpixel Segmentation Methods Influence Deforestation Image Classification?

链接: https://arxiv.org/abs/2510.04645
作者: Hugo Resende,Fabio A. Faria,Eduardo B. Neto,Isabela Borlido,Victor Sundermann,Silvio Jamil F. Guimarães,Álvaro L. Fazenda
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 15 pages, 3 figures, paper accepted to present at CIARP 2025

点击查看摘要

[CV-51] Social Agent : Mastering Dyadic Nonverbal Behavior Generation via Conversational LLM Agents SIGGRAPH

链接: https://arxiv.org/abs/2510.04637
作者: Zeyi Zhang,Yanju Zhou,Heyuan Yao,Tenglong Ao,Xiaohang Zhan,Libin Liu
机构: Peking University (北京大学); Tencent (腾讯)
类目: Graphics (cs.GR); Computer Vision and Pattern Recognition (cs.CV)
备注: SIGGRAPH ASIA 2025 (Conference Track); Project page: this https URL

点击查看摘要

[CV-52] SFANet: Spatial-Frequency Attention Network for Deepfake Detection

链接: https://arxiv.org/abs/2510.04630
作者: Vrushank Ahire,Aniruddh Muley,Shivam Zample,Siddharth Verma,Pranav Menon,Surbhi Madan,Abhinav Dhall
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Multimedia (cs.MM)
备注:

点击查看摘要

[CV-53] A Spatial-Spectral-Frequency Interactive Network for Multimodal Remote Sensing Classification

链接: https://arxiv.org/abs/2510.04628
作者: Hao Liu,Yunhao Gao,Wei Li,Mingyang Zhang,Maoguo Gong,Lorenzo Bruzzone
机构: University of Trento (特伦托大学); Beijing Institute of Technology (北京理工大学); Xidian University (西安电子科技大学); Inner Mongolia Normal University (内蒙古师范大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-54] Pathology-CoT: Learning Visual Chain-of-Thought Agent from Expert Whole Slide Image Diagnosis Behavior

链接: https://arxiv.org/abs/2510.04587
作者: Sheng Wang,Ruiming Wu,Charles Herndon,Yihang Liu,Shunsuke Koga,Jeanne Shen,Zhi Huang
机构: University of Pennsylvania (宾夕法尼亚大学); University of California at San Francisco (加州大学旧金山分校); Stanford University (斯坦福大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-55] SONA: Learning Conditional Unconditional and Mismatching-Aware Discriminator

链接: https://arxiv.org/abs/2510.04576
作者: Yuhta Takida,Satoshi Hayakawa,Takashi Shibuya,Masaaki Imaizumi,Naoki Murata,Bac Nguyen,Toshimitsu Uesaka,Chieh-Hsin Lai,Yuki Mitsufuji
机构: Sony AI; Sony Group Corporation; The University of Tokyo
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (stat.ML)
备注: 24 pages with 9 figures

点击查看摘要

[CV-56] Conditional Representation Learning for Customized Tasks

链接: https://arxiv.org/abs/2510.04564
作者: Honglin Liu,Chao Sun,Peng Hu,Yunfan Li,Xi Peng
机构: Sichuan University (四川大学); Chinese Academy of Sciences (中国科学院)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-57] Fast Witness Persistence for MRI Volumes via Hybrid Landmarking

链接: https://arxiv.org/abs/2510.04553
作者: Jorge Leonardo Ruiz Williams
机构: 未知
类目: Computational Geometry (cs.CG); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:

点击查看摘要

[CV-58] Post-training quantization of vision encoders needs prefixing registers

链接: https://arxiv.org/abs/2510.04547
作者: Seunghyeon Kim,Jinho Kim,Taesun Yeom,Wonpyo Park,Kyuyeun Kim,Jaeho Lee
机构: 未知
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-59] C3Editor: Achieving Controllable Consistency in 2D Model for 3D Editing

链接: https://arxiv.org/abs/2510.04539
作者: Zeng Tao,Zheng Ding,Zeyuan Chen,Xiang Zhang,Leizhi Li,Zhuowen Tu
机构: Fudan University (复旦大学); UC San Diego (加州大学圣地亚哥分校)
类目: Graphics (cs.GR); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-60] 3Dify: a Framework for Procedural 3D-CG Generation Assisted by LLM s Using MCP and RAG

链接: https://arxiv.org/abs/2510.04536
作者: Shun-ichiro Hayashi,Daichi Mukunoki,Tetsuya Hoshino,Satoshi Ohshima,Takahiro Katagiri
机构: 未知
类目: Graphics (cs.GR); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-61] AG:Tangential Amplifying Guidance for Hallucination-Resistant Diffusion Sampling

链接: https://arxiv.org/abs/2510.04533
作者: Hyunmin Cho,Donghoon Ahn,Susung Hong,Jee Eun Kim,Seungryong Kim,Kyong Hwan Jin
机构: Korea University (韩国大学); University of California, Berkeley (加州大学伯克利分校); University of Washington (华盛顿大学); KAIST AI (韩国科学技术院人工智能)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 16 pages, 9 figures, 5 tables

点击查看摘要

[CV-62] Real-time Prediction of Urban Sound Propagation with Conditioned Normalizing Flows

链接: https://arxiv.org/abs/2510.04510
作者: Achim Eckerle,Martin Spitznagel,Janis Keuper
机构: Stralsund University (斯特拉尔松德大学); IMLA, Offenburg University (Offenburg 大学智能机器学习研究所)
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-63] Asynchronous Denoising Diffusion Models for Aligning Text-to-Image Generation

链接: https://arxiv.org/abs/2510.04504
作者: Zijing Hu,Yunze Tong,Fengda Zhang,Junkun Yuan,Jun Xiao,Kun Kuang
机构: Zhejiang University (浙江大学); Nanyang Technological University (南洋理工大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 22 pages, 11 figures, 5 tables

点击查看摘要

[CV-64] BStar-Edit: From Image Editing Pattern Shifting to Consistency Enhancement

链接: https://arxiv.org/abs/2510.04483
作者: Hao Fang,Zechao Zhan,Weixin Feng,Ziwei Huang,XuBin Li,Tiezheng Ge
机构: Taobao & Tmall Group of Alibaba(淘宝与天猫集团)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-65] VaseVQA-3D: Benchmarking 3D VLMs on Ancient Greek Pottery

链接: https://arxiv.org/abs/2510.04479
作者: Nonghai Zhang,Zeyu Zhang,Jiazi Wang,Yang Zhao,Hao Tang
机构: Peking University (北京大学); Beijing Jiaotong University (北京交通大学); La Trobe University (拉特罗布大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-66] SPEGNet: Synergistic Perception-Guided Network for Camouflaged Object Detection

链接: https://arxiv.org/abs/2510.04472
作者: Baber Jan,Saeed Anwar,Aiman H. El-Maleh,Abdul Jabbar Siddiqui,Abdul Bais
机构: King Fahd University of Petroleum and Minerals (国王法赫德石油矿产大学); The University of Western Australia (西澳大利亚大学); University of Regina (里贾纳大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Image and Video Processing (eess.IV)
备注:

点击查看摘要

[CV-67] REAR: Rethinking Visual Autoregressive Models via Generator-Tokenizer Consistency Regularization

链接: https://arxiv.org/abs/2510.04450
作者: Qiyuan He,Yicong Li,Haotian Ye,Jinghao Wang,Xinyao Liao,Pheng-Ann Heng,Stefano Ermon,James Zou,Angela Yao
机构: National University of Singapore (新加坡国立大学); Stanford University (斯坦福大学); The Chinese University of Hong Kong (香港中文大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 27 pages, 23 figures, 5 tables

点击查看摘要

[CV-68] A.I.R.: Enabling Adaptive Iterative and Reasoning -based Frame Selection For Video Question Answering

链接: https://arxiv.org/abs/2510.04428
作者: Yuanhao Zou,Shengji Jin,Andong Deng,Youpeng Zhao,Jun Wang,Chen Chen
机构: University of Central Florida (中佛罗里达大学); Weill Cornell Medicine (威尔康奈尔医学院)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-69] CodeFormer: Blind Face Restoration Using Deformable Registration and Deep Metric Learning

链接: https://arxiv.org/abs/2510.04410
作者: Venkata Bharath Reddy Reddem,Akshay P Sarashetti,Ranjith Merugu,Amit Satish Unde
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-70] Your Vision-Language Model Cant Even Count to 20: Exposing the Failures of VLMs in Compositional Counting

链接: https://arxiv.org/abs/2510.04401
作者: Xuyang Guo,Zekai Huang,Zhenmei Shi,Zhao Song,Jiahao Zhang
机构: Guilin University of Electronic Technology (桂林电子科技大学); The Ohio State University (俄亥俄州立大学); University of Wisconsin-Madison (威斯康星大学麦迪逊分校); University of California, Berkeley (加州大学伯克利分校)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[CV-71] Diffusion2: Dual Diffusion Model with Uncertainty-Aware Adaptive Noise for Momentary Trajectory Prediction

链接: https://arxiv.org/abs/2510.04365
作者: Yuhao Luo,Yuang Zhang,Kehua Chen,Xinyu Zheng,Shucheng Zhang,Sikai Chen,Yinhai Wang
机构: University of Wisconsin-Madison (威斯康星大学麦迪逊分校); University of Washington (华盛顿大学); Tongji University (同济大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 13 pages, 7 figures, 3 tables

点击查看摘要

[CV-72] RAP: 3D Rasterization Augmented End-to-End Planning

链接: https://arxiv.org/abs/2510.04333
作者: Lan Feng,Yang Gao,Eloi Zablocki,Quanyi Li,Wuyang Li,Sichao Liu,Matthieu Cord,Alexandre Alahi
机构: EPFL(瑞士联邦理工学院); Valeo.ai(法雷奥人工智能); Sorbonne Université(索邦大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:

点击查看摘要

[CV-73] DoRAN: Stabilizing Weight-Decomposed Low-Rank Adaptation via Noise Injection and Auxiliary Networks

链接: https://arxiv.org/abs/2510.04331
作者: Nghiem T. Diep,Hien Dang,Tuan Truong,Tan Dinh,Huy Nguyen,Nhat Ho
机构: 未知
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: Nghiem T. Diep, Hien Dang, and Tuan Truong contributed equally to this work

点击查看摘要

[CV-74] GenAR: Next-Scale Autoregressive Generation for Spatial Gene Expression Prediction

链接: https://arxiv.org/abs/2510.04315
作者: Jiarui Ouyang,Yihui Wang,Yihang Gao,Yingxue Xu,Shu Yang,Hao Chen
机构: The Hong Kong University of Science and Technology (香港科技大学); Shenzhen Loop Area Institute (深圳环区研究院)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-75] CARE-PD: A Multi-Site Anonymized Clinical Dataset for Parkinsons Disease Gait Assessment NEURIPS2025

链接: https://arxiv.org/abs/2510.04312
作者: Vida Adeli,Ivan Klabucar,Javad Rajabi,Benjamin Filtjens,Soroush Mehraban,Diwei Wang,Hyewon Seo,Trung-Hieu Hoang,Minh N. Do,Candice Muller,Claudia Oliveira,Daniel Boari Coelho,Pieter Ginis,Moran Gilat,Alice Nieuwboer,Joke Spildooren,Lucas Mckay,Hyeokhyen Kwon,Gari Clifford,Christine Esper,Stewart Factor,Imari Genias,Amirhossein Dadashzadeh,Leia Shum,Alan Whone,Majid Mirmehdi,Andrea Iaboni,Babak Taati
机构: University of Toronto (多伦多大学); Vector Institute (向量研究所); KITE Research Institute-UHN (KITE 研究所-UHN); University of Strasbourg (斯特拉斯堡大学); University Hospitals of Strasbourg (斯特拉斯堡大学医院); University of Illinois Urbana-Champaign (伊利诺伊大学厄巴纳-香槟分校); Federal University of ABC (ABC联邦大学); KU Leuven (鲁汶大学); Hasselt University (哈塞尔特大学); Emory University (埃默里大学); University of Bristol (布里斯托大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at the Thirty-Ninth Conference on Neural Information Processing Systems (NeurIPS 2025)

点击查看摘要

[CV-76] ChronoEdit: Towards Temporal Reasoning for Image Editing and World Simulation

链接: https://arxiv.org/abs/2510.04290
作者: Jay Zhangjie Wu,Xuanchi Ren,Tianchang Shen,Tianshi Cao,Kai He,Yifan Lu,Ruiyuan Gao,Enze Xie,Shiyi Lan,Jose M. Alvarez,Jun Gao,Sanja Fidler,Zian Wang,Huan Ling
机构: NVIDIA(英伟达); University of Toronto(多伦多大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL

点击查看摘要

[CV-77] Flexible and Efficient Spatio-Temporal Transformer for Sequential Visual Place Recognition

链接: https://arxiv.org/abs/2510.04282
作者: Yu Kiu(Idan)Lau,Chao Chen,Ge Jin,Chen Feng
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 8 pages, 6 figures

点击查看摘要

[CV-78] Concept-Based Masking: A Patch-Agnostic Defense Against Adversarial Patch Attacks NEURIPS

链接: https://arxiv.org/abs/2510.04245
作者: Ayushi Mehrotra,Derek Peng,Dipkamal Bhusal,Nidhi Rastogi
机构: California Institute of Technology (加州理工学院); University of California, Berkeley (加州大学伯克利分校); Rochester Institute of Technology (罗切斯特理工学院)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: neurips workshop

点击查看摘要

[CV-79] he best performance in the CARE 2025 – Liver Task (LiSeg-Contrast): Contrast-Aware Semi-Supervised Segmentation with Domain Generalization and Test-Time Adaptation

链接: https://arxiv.org/abs/2510.04243
作者: Jincan Lou,Jingkun Chen,Haoquan Li,Hang Li,Wenjian Huang,Weihua Chen,Fan Wang,Jianguo Zhang
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 11 pages, 3 figures

点击查看摘要

[CV-80] Scaling Sequence-to-Sequence Generative Neural Rendering

链接: https://arxiv.org/abs/2510.04236
作者: Shikun Liu,Kam Woh Ng,Wonbong Jang,Jiadong Guo,Junlin Han,Haozhe Liu,Yiannis Douratsos,Juan C. Pérez,Zijian Zhou,Chi Phung,Tao Xiang,Juan-Manuel Pérez-Rúa
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL

点击查看摘要

[CV-81] Detection of retinal diseases using an accelerated reused convolutional network

链接: https://arxiv.org/abs/2510.04232
作者: Amin Ahmadi Kasani,Hedieh Sajedi
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:

点击查看摘要

[CV-82] A Recursive Pyramidal Algorithm for Solving the Image Registration Problem

链接: https://arxiv.org/abs/2510.04231
作者: Stefan Dirnstorfer
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-83] MASC: Boosting Autoregressive Image Generation with a Manifold-Aligned Semantic Clustering

链接: https://arxiv.org/abs/2510.04220
作者: Lixuan He,Shikang Zheng,Linfeng Zhang
机构: Shanghai Jiao Tong University (上海交通大学); Tsinghua University (清华大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[CV-84] World-To-Image: Grounding Text-to-Image Generation with Agent -Driven World Knowledge

链接: https://arxiv.org/abs/2510.04201
作者: Moo Hyun Son,Jintaek Oh,Sun Bin Mun,Jaechul Roh,Sehyun Choi
机构: The Hong Kong University of Science and Technology (香港科技大学); Georgia Institute of Technology (佐治亚理工学院); University of Massachusetts Amherst (马萨诸塞大学阿默斯特分校); TwelveLabs
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[CV-85] Let Features Decide Their Own Solvers: Hybrid Feature Caching for Diffusion Transformers

【速读】：该论文旨在解决扩散模型（Diffusion Models）在图像和视频生成中因迭代采样过程导致的计算效率瓶颈问题，尤其是Transformer前向传播在每个时间步的高开销。现有特征缓存（feature caching）方法通常采用统一策略对所有特征维度进行处理，忽略了不同维度间动态行为的异质性。为此，作者提出HyCa框架，其核心创新在于将隐藏特征演化建模为跨维度的常微分方程（ODE）混合系统，并设计了一种受混合ODE求解器启发的维度感知缓存策略，从而实现更精准、高效的特征复用。该方案无需重新训练即可在多种模型上实现显著加速（如FLUX、HunyuanVideo等），且保持近无损的生成质量。

链接: https://arxiv.org/abs/2510.04188
作者: Shikang Zheng,Guantao Chen,Qinming Zhou,Yuqi Lin,Lixuan He,Chang Zou,Peiliang Cai,Jiacheng Liu,Linfeng Zhang
机构: Shanghai Jiao Tong University (上海交通大学); South China University of Technology (华南理工大学); Tsinghua University (清华大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

Abstract:Diffusion Transformers offer state-of-the-art fidelity in image and video synthesis, but their iterative sampling process remains a major bottleneck due to the high cost of transformer forward passes at each timestep. To mitigate this, feature caching has emerged as a training-free acceleration technique that reuses or forecasts hidden representations. However, existing methods often apply a uniform caching strategy across all feature dimensions, ignoring their heterogeneous dynamic behaviors. Therefore, we adopt a new perspective by modeling hidden feature evolution as a mixture of ODEs across dimensions, and introduce HyCa, a Hybrid ODE solver inspired caching framework that applies dimension-wise caching strategies. HyCa achieves near-lossless acceleration across diverse domains and models, including 5.55 times speedup on FLUX, 5.56 times speedup on HunyuanVideo, 6.24 times speedup on Qwen-Image and Qwen-Image-Edit without retraining.
zh

[CV-86] From Segments to Concepts: Interpretable Image Classification via Concept-Guided Segmentation

【速读】：该论文旨在解决深度神经网络在计算机视觉任务中因决策过程缺乏透明性而导致的可解释性不足问题，尤其是在安全关键场景下，模型可能依赖不可靠或误导性特征，从而影响鲁棒性和解释有效性。其核心解决方案是提出SEG-MIL-CBM框架，该框架将概念引导的图像分割与基于注意力机制的多实例学习（Multiple Instance Learning, MIL）相结合，将每个分割区域视为一个实例，通过聚合跨区域证据来实现对高阶语义概念的推理。该方法无需概念标注或分组信息即可生成空间上对齐、概念层面的解释，同时显著提升模型在虚假相关性（spurious correlations）、输入扰动和大规模基准测试中的鲁棒性能。

链接: https://arxiv.org/abs/2510.04180
作者: Ran Eisenberg,Amit Rozner,Ethan Fetaya,Ofir Lindenbaum
机构: Bar-Ilan University (巴伊兰大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:

点击查看摘要

Abstract:Deep neural networks have achieved remarkable success in computer vision; however, their black-box nature in decision-making limits interpretability and trust, particularly in safety-critical applications. Interpretability is crucial in domains where errors have severe consequences. Existing models not only lack transparency but also risk exploiting unreliable or misleading features, which undermines both robustness and the validity of their explanations. Concept Bottleneck Models (CBMs) aim to improve transparency by reasoning through human-interpretable concepts. Still, they require costly concept annotations and lack spatial grounding, often failing to identify which regions support each concept. We propose SEG-MIL-CBM, a novel framework that integrates concept-guided image segmentation into an attention-based multiple instance learning (MIL) framework, where each segmented region is treated as an instance and the model learns to aggregate evidence across them. By reasoning over semantically meaningful regions aligned with high-level concepts, our model highlights task-relevant evidence, down-weights irrelevant cues, and produces spatially grounded, concept-level explanations without requiring annotations of concepts or groups. SEG-MIL-CBM achieves robust performance across settings involving spurious correlations (unintended dependencies between background and label), input corruptions (perturbations that degrade visual quality), and large-scale benchmarks, while providing transparent, concept-level explanations.
zh

[CV-87] BLADE: Bias-Linked Adaptive DEbiasing

链接: https://arxiv.org/abs/2510.04174
作者: Piyush Arora,Navlika Singh,Vasubhya Diwan,Pratik Mazumder
机构: Indian Institute of Technology Jodhpur (印度理工学院贾德普尔分校)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: The authors have contributed equally

点击查看摘要

[CV-88] Learning from All: Concept Alignment for Autonomous Distillation from Multiple Drifting MLLM s

链接: https://arxiv.org/abs/2510.04142
作者: Xiaoyu Yang,Jie Lu,En Yu
机构: Australian Artificial Intelligence Institute (澳大利亚人工智能研究所); University of Technology Sydney (悉尼科技大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[CV-89] Learning-Based Hashing for ANN Search: Foundations and Early Advances

链接: https://arxiv.org/abs/2510.04127
作者: Sean Moran
机构: 未知
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:

点击查看摘要

[CV-90] Joint Learning of Pose Regression and Denoising Diffusion with Score Scaling Sampling for Category-level 6D Pose Estimation

链接: https://arxiv.org/abs/2510.04125
作者: Seunghyun Lee,Tae-Kyun Kim
机构: KAIST
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-91] Learning Efficient Meshflow and Optical Flow from Event Cameras

链接: https://arxiv.org/abs/2510.04111
作者: Xinglong Luo,Ao Luo,Kunming Luo,Zhengning Wang,Ping Tan,Bing Zeng,Shuaicheng Liu
机构: Institute of Image Processing, School of Information and Communication Engineering, University of Electronic Science and Technology of China (电子科技大学信息与通信工程学院图像处理研究所); Yibin Institute of UESTC (宜宾研究院); School of Computing and Artificial Intelligence, Southwest Jiaotong University (西南交通大学计算机与人工智能学院); Department of Electronic and Computer Engineering, The Hong Kong University of Science and Technology (香港科技大学电子与计算机工程系)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by TPAMI 2025

点击查看摘要

[CV-92] OPO-Bench: An Open-Source Topological Mapping Evaluation Framework with Quantifiable Perceptual Aliasing

链接: https://arxiv.org/abs/2510.04100
作者: Jiaming Wang,Diwen Liu,Jizhuo Chen,Harold Soh
机构: National University of Singapore (新加坡国立大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Jiaming Wang, Diwen Liu, and Jizhuo Chen contributed equally

点击查看摘要

[CV-93] Using predefined vector systems as latent space configuration for neural network supervised training on data with arbitrarily large number of classes

链接: https://arxiv.org/abs/2510.04090
作者: Nikita Gabdullin
机构: Joint Stock "Research and production company “Kryptonite”
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 28 pages, 12 figures, 10 tables, 12 equations, 1 algorithm

点击查看摘要

[CV-94] Diffusion Low Rank Hybrid Reconstruction for Sparse View Medical Imaging

链接: https://arxiv.org/abs/2510.04069
作者: Zongyin Deng,Qing Zhou,Yuhao Fang,Zijian Wang,Yao Lu,Ye Zhang,Chun Li
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-95] QuantDemoire: Quantization with Outlier Aware for Image Demoiréing

链接: https://arxiv.org/abs/2510.04066
作者: Zheng Chen,Kewei Zhang,Xiaoyang Liu,Weihang Zhang,Mengfan Wang,Yifan Fu,Yulun Zhang
机构: Shanghai Jiao Tong University (上海交通大学); Central Media Technology Institute, Huawei (华为中央媒体技术研究院)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Code is available at: this https URL

点击查看摘要

[CV-96] Ordinal Encoding as a Regularizer in Binary Loss for Solar Flare Prediction ICDM

链接: https://arxiv.org/abs/2510.04063
作者: Chetraj Pandey,Jinsu Hong,Anli Ji,Rafal A. Angryk,Berkay Aydin
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Solar and Stellar Astrophysics (astro-ph.SR)
备注: This is a preprint submitted to ICDM Workshop (SABID 2025). 6 pages, 2 Figures

点击查看摘要

[CV-97] MetaFind: Scene-Aware 3D Asset Retrieval for Coherent Metaverse Scene Generation NEURIPS2025

链接: https://arxiv.org/abs/2510.04057
作者: Zhenyu Pan,Yucheng Lu,Han Liu
机构: Northwestern University (西北大学); New York University (纽约大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: The Thirty-Ninth Annual Conference on Neural Information Processing Systems (NeurIPS 2025)

点击查看摘要

[CV-98] Quantization Range Estimation for Convolutional Neural Networks

链接: https://arxiv.org/abs/2510.04044
作者: Bingtao Yang,Yujia Wang,Mengzhi Jiao,Hongwei Huo
机构: Xidian University (西安电子科技大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 11 pages, 5 tables, research report

点击查看摘要

[CV-99] textscGUI-Spotlight: Adaptive Iterative Focus Refinement for Enhanced GUI Visual Grounding

链接: https://arxiv.org/abs/2510.04039
作者: Bin Lei,Nuo Xu,Ali Payani,Mingyi Hong,Chunhua Liao,Yu Cao,Caiwen Ding
机构: University of Minnesota (明尼苏达大学); Cisco Research (思科研究院); Lawrence Livermore National Labs (劳伦斯利弗莫尔国家实验室)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[CV-100] Prompt-to-Prompt: Text-Based Image Editing Via Cross-Attention Mechanisms – The Research of Hyperparameters and Novel Mechanisms to Enhance Existing Frameworks

【速读】：该论文旨在解决当前prompt-to-prompt图像编辑框架中因超参数设置不当导致的生成结果不稳定问题，例如头发颜色等细节变化不一致。其解决方案的关键在于通过系统性优化超参数并引入改进机制：首先深入研究“词替换”（word swap）方法；其次提出“注意力重加权”（attention re-weight method）以提升模型对不同提示词的适应能力；最后构建CL P2P框架，有效缓解循环一致性（cycle inconsistency）等现有局限。该工作揭示了超参数与神经网络注意力机制之间的交互关系，从而显著提升了图像编辑的精度与可靠性。

链接: https://arxiv.org/abs/2510.04034
作者: Linn Bieske,Carla Lorente
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

Abstract:Recent advances in image editing have shifted from manual pixel manipulation to employing deep learning methods like stable diffusion models, which now leverage cross-attention mechanisms for text-driven control. This transition has simplified the editing process but also introduced variability in results, such as inconsistent hair color changes. Our research aims to enhance the precision and reliability of prompt-to-prompt image editing frameworks by exploring and optimizing hyperparameters. We present a comprehensive study of the “word swap” method, develop an “attention re-weight method” for better adaptability, and propose the “CL P2P” framework to address existing limitations like cycle inconsistency. This work contributes to understanding and improving the interaction between hyperparameter settings and the architectural choices of neural network models, specifically their attention mechanisms, which significantly influence the composition and quality of the generated images.
zh

[CV-101] Enhancing Fake News Video Detection via LLM -Driven Creative Process Simulation CIKM2025

链接: https://arxiv.org/abs/2510.04024
作者: Yuyan Bu,Qiang Sheng,Juan Cao,Shaofei Wang,Peng Qi,Yuhui Shi,Beizhe Hu
机构: Media Synthesis and Forensics Lab, Institute of Computing Technology, Chinese Academy of Sciences (中国科学院计算技术研究所媒体合成与取证实验室); University of Chinese Academy of Sciences (中国科学院大学); National University of Singapore (新加坡国立大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
备注: ACM CIKM 2025

点击查看摘要

[CV-102] Video-in-the-Loop: Span-Grounded Long Video QA with Interleaved Reasoning

链接: https://arxiv.org/abs/2510.04022
作者: Chendong Wang,Donglin Bai,Yifan Yang,Xiao Jin,Anlan Zhang,Rui Wang,Shiqi Jiang,Yuqing Yang,Hao Wu,Qi Dai,Chong Luo,Ting Cao,Lili Qiu,Suman Banerjee
机构: University of Wisconsin–Madison (威斯康星大学麦迪逊分校); Microsoft Research Asia (微软亚洲研究院); Columbia University (哥伦比亚大学); University of Southern California (南加州大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-103] Fit Pixels Get Labels: Meta-learned Implicit Networks for Image Segmentation MICCAI2025

链接: https://arxiv.org/abs/2510.04021
作者: Kushal Vyas,Ashok Veeraraghavan,Guha Balakrishnan
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: MICCAI 2025 (oral). Final peer-reviewed copy accessible at publisher DOI this https URL . Project page, this https URL

点击查看摘要

[CV-104] Keep It on a Leash: Controllable Pseudo-label Generation Towards Realistic Long-Tailed Semi-Supervised Learning NEURIPS2025

链接: https://arxiv.org/abs/2510.03993
作者: Yaxin Hou,Bo Han,Yuheng Jia,Hui Liu,Junhui Hou
机构: Southeast University (东南大学); Saint Francis University (圣弗朗西斯大学); City University of Hong Kong (香港城市大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: The paper is accepted by NeurIPS 2025

点击查看摘要

[CV-105] Use of Quadcopter Wakes to Supplement Strawberry Pollination

链接: https://arxiv.org/abs/2510.03974
作者: Sadie Cutler,Ben DeFay,Scott McArt,Kirstin Petersen
机构: Cornell Institute for Digital Agriculture (Cornell大学数字农业研究所); Sibley School of Mechanical and Aerospace Engineering, Cornell University (康奈尔大学机械与航空航天工程系); School of Electrical and Computer Engineering, Cornell University (康奈尔大学电气与计算机工程学院); College of Agriculture and Life Science, Cornell University (康奈尔大学农业与生命科学学院)
类目: ystems and Control (eess.SY); Computer Vision and Pattern Recognition (cs.CV)
备注: 7 pages, 7 figures

点击查看摘要

[CV-106] Harnessing Synthetic Preference Data for Enhancing Temporal Understanding of Video-LLM s

链接: https://arxiv.org/abs/2510.03955
作者: Sameep Vani,Shreyas Jena,Maitreya Patel,Chitta Baral,Somak Aditya,Yezhou Yang
机构: Arizona State University (亚利桑那州立大学); Indian Institute of Technology, Kharagpur (印度理工学院克哈格帕尔分校)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 17 pages, 9 figures, 6 tables. Presents TimeWarp, a synthetic preference data framework to improve temporal understanding in Video-LLMs, showing consistent gains across seven benchmarks. Includes supplementary material in the Appendix

点击查看摘要

[CV-107] alking Tennis: Language Feedback from 3D Biomechanical Action Recognition

链接: https://arxiv.org/abs/2510.03921
作者: Arushi Dashore,Aryan Anumala,Emily Hui,Olivia Yang
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注: 10 pages, 4 figures, 2 tables

点击查看摘要

[CV-108] OpenFLAME: Federated Visual Positioning System to Enable Large-Scale Augmented Reality Applications

链接: https://arxiv.org/abs/2510.03915
作者: Sagar Bharadwaj,Harrison Williams,Luke Wang,Michael Liang,Tao Jin,Srinivasan Seshan,Anthony Rowe
机构: Carnegie Mellon University (卡内基梅隆大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Distributed, Parallel, and Cluster Computing (cs.DC); Robotics (cs.RO)
备注:

点击查看摘要

[CV-109] Generating Human Motion Videos using a Cascaded Text-to-Video Framework

【速读】：该论文旨在解决通用人类运动视频生成（general human motion video generation）中存在的挑战，尤其是现有视频扩散模型（Video Diffusion Models, VDMs）在跨模态对齐、条件控制和视角一致性方面的不足。当前多数方法受限于图像到视频的设定或特定领域（如舞蹈视频），难以实现文本驱动的多样化人类动作生成。其解决方案的关键在于提出CAMEO框架——一个级联式架构，通过精心设计的组件无缝融合文本到动作（Text-to-Motion, T2M）模型与条件VDM，从而在训练和推理阶段均有效缓解次优因素。核心创新包括：1）对文本提示与视觉条件进行系统性预处理以增强模型对齐能力；2）引入相机感知条件模块（camera-aware conditioning module），自动选择与输入文本语义一致的摄像机视角，提升视频时空一致性并减少人工干预。

链接: https://arxiv.org/abs/2510.03909
作者: Hyelin Nam,Hyojun Go,Byeongjun Park,Byung-Hoon Kim,Hyungjin Chung
机构: EverEx; University of Michigan (密歇根大学); ETH Zurich (苏黎世联邦理工学院); Yonsei University (延世大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 18 pages, 7 figures, Project Page: this https URL

点击查看摘要

Abstract:Human video generation is becoming an increasingly important task with broad applications in graphics, entertainment, and embodied AI. Despite the rapid progress of video diffusion models (VDMs), their use for general-purpose human video generation remains underexplored, with most works constrained to image-to-video setups or narrow domains like dance videos. In this work, we propose CAMEO, a cascaded framework for general human motion video generation. It seamlessly bridges Text-to-Motion (T2M) models and conditional VDMs, mitigating suboptimal factors that may arise in this process across both training and inference through carefully designed components. Specifically, we analyze and prepare both textual prompts and visual conditions to effectively train the VDM, ensuring robust alignment between motion descriptions, conditioning signals, and the generated videos. Furthermore, we introduce a camera-aware conditioning module that connects the two stages, automatically selecting viewpoints aligned with the input text to enhance coherence and reduce manual intervention. We demonstrate the effectiveness of our approach on both the MovieGen benchmark and a newly introduced benchmark tailored to the T2M-VDM combination, while highlighting its versatility across diverse use cases.
zh

[CV-110] From Filters to VLMs: Benchmarking Defogging Methods through Object Detection and Segmentation Performance

链接: https://arxiv.org/abs/2510.03906
作者: Ardalan Aryashad,Parsa Razmara,Amin Mahjoub,Seyedarmin Azizi,Mahdi Salmani,Arad Firouzkouhi
机构: University of Southern California (南加州大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-111] Zero-Shot Fine-Grained Image Classification Using Large Vision-Language Models EMNLP2025

链接: https://arxiv.org/abs/2510.03903
作者: Md. Atabuzzaman,Andrew Zhang,Chris Thomas
机构: Virginia Tech (弗吉尼亚理工大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to EMNLP 2025 Findings

点击查看摘要

[CV-112] Bridge Thinking and Acting: Unleashing Physical Potential of VLM with Generalizable Action Expert

链接: https://arxiv.org/abs/2510.03896
作者: Mingyu Liu,Zheng Huang,Xiaoyi Lin,Muzhi Zhu,Canyu Zhao,Zongze Du,Yating Wang,Haoyi Zhu,Hao Chen,Chunhua Shen
机构: Zhejiang University (浙江大学); Shanghai Artificial Intelligence Laboratory (上海人工智能实验室)
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:

点击查看摘要

[CV-113] NoTVLA: Narrowing of Dense Action Trajectories for Generalizable Robot Manipulation

链接: https://arxiv.org/abs/2510.03895
作者: Zheng Huang,Mingyu Liu,Xiaoyi Lin,Muzhi Zhu,Canyu Zhao,Zongze Du,Xiaoman Li,Yiduo Jia,Hao Zhong,Hao Chen,Chunhua Shen
机构: Zhejiang University (浙江大学)
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-114] Exploring Instruction Data Quality for Explainable Image Quality Assessment

链接: https://arxiv.org/abs/2510.03880
作者: Yunhao Li,Sijing Wu,Huiyu Duan,Yucheng Zhu,Qi Jia,Guangtao Zhai
机构: Shanghai Jiao Tong University (上海交通大学); Shanghai Artificial Intelligence Laboratory (上海人工智能实验室)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-115] Multi-Modal Oral Cancer Detection Using Weighted Ensemble Convolutional Neural Networks

链接: https://arxiv.org/abs/2510.03878
作者: Ajo Babu George,Sreehari J R Ajo Babu George,Sreehari J R Ajo Babu George,Sreehari J R
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[CV-116] Skin Lesion Classification Based on ResNet-50 Enhanced With Adaptive Spatial Feature Fusion

链接: https://arxiv.org/abs/2510.03876
作者: Runhao Liu,Ziming Chen,Peng Zhang
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-117] DHQA-4D: Perceptual Quality Assessment of Dynamic 4D Digital Human

链接: https://arxiv.org/abs/2510.03874
作者: Yunhao Li,Sijing Wu,Yucheng Zhu,Huiyu Duan,Zicheng Zhang,Guangtao Zhai
机构: Institute of Image Communication and Network Engineering, Shanghai Jiao Tong University (上海交通大学图像通信与网络工程研究所); USC-SJTU Institute of Cultural and Creative Industry, Shanghai Jiao Tong University (上海交通大学南加州大学文化创意产业联合研究院)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-118] PoseGaze-AHP: A Knowledge-Based 3D Dataset for AI-Driven Ocular and Postural Diagnosis

链接: https://arxiv.org/abs/2510.03873
作者: Saja Al-Dabet,Sherzod Turaev,Nazar Zaki,Arif O. Khan,Luai Eldweik
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: This is a preprint version of a manuscript under review. All rights reserved by the authors

点击查看摘要

[CV-119] SDAKD: Student Discriminator Assisted Knowledge Distillation for Super-Resolution Generative Adversarial Networks

链接: https://arxiv.org/abs/2510.03870
作者: Nikolaos Kaparinos,Vasileios Mezaris
机构: CERTH-ITI (CERTH-ITI)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Under review

点击查看摘要

[CV-120] Exploring the Challenge and Value of Deep Learning in Automated Skin Disease Diagnosis

链接: https://arxiv.org/abs/2510.03869
作者: Runhao Liu,Ziming Chen,Peng Zhang
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-121] Cross-View Open-Vocabulary Object Detection in Aerial Imagery

链接: https://arxiv.org/abs/2510.03858
作者: Jyoti Kini,Rohit Gupta,Mubarak Shah
机构: Center for Research in Computer Vision, University of Central Florida (中央佛罗里达大学计算机视觉研究中心)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-122] Optimized Minimal 4D Gaussian Splatting

链接: https://arxiv.org/abs/2510.03857
作者: Minseo Lee,Byeonghyeon Lee,Lucas Yunkyu Lee,Eunsoo Lee,Sangmin Kim,Seunghyeon Song,Joo Chan Lee,Jong Hwan Ko,Jaesik Park,Eunbyung Park
机构: Yonsei University (延世大学); Seoul National University (首尔国立大学); POSTECH (浦项工科大学); Sungkyunkwan University (成均馆大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 17 pages, 8 figures

点击查看摘要

[CV-123] UGround: Towards Unified Visual Grounding with Unrolled Transformers

链接: https://arxiv.org/abs/2510.03853
作者: Rui Qian,Xin Yin,Chuanhang Deng,Zhiyuan Peng,Jian Xiong,Wei Zhai,Dejing Dou
机构: Fudan University (复旦大学); Zhejiang University (浙江大学); Shanghai Jiao Tong University (上海交通大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: this https URL

点击查看摘要

[CV-124] Mirag e: Unveiling Hidden Artifacts in Synthetic Images with Large Vision-Language Models ACM-MM’25

链接: https://arxiv.org/abs/2510.03840
作者: Pranav Sharma,Shivank Garg,Durga Toshniwal
机构: Indian Institute of Technology Roorkee(印度理工学院鲁尔基分校)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: ACM MM’25, MALLM Workshop

点击查看摘要

[CV-125] Joint Neural SDF Reconstruction and Semantic Segmentation for CAD Models

链接: https://arxiv.org/abs/2510.03837
作者: Shen Fan,Przemyslaw Musialski
机构: 未知
类目: Graphics (cs.GR); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-126] LIBERO-PRO: Towards Robust and Fair Evaluation of Vision-Language-Action Models Beyond Memorization

链接: https://arxiv.org/abs/2510.03827
作者: Xueyang Zhou,Yangming Xu,Guiyao Tie,Yongchao Chen,Guowen Zhang,Duanfeng Chu,Pan Zhou,Lichao Sun
机构: Huazhong University of Science and Technology (华中科技大学); Harvard University (哈佛大学); Massachusetts Institute of Technology (麻省理工学院); Wuhan University of Technology (武汉理工大学); Lehigh University (利海大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 12 pages,7 figures, 5 tables

点击查看摘要

[CV-127] Contrastive-SDE: Guiding Stochastic Differential Equations with Contrastive Learning for Unpaired Image-to-Image Translation

链接: https://arxiv.org/abs/2510.03821
作者: Venkata Narendra Kotyada,Revanth Eranki,Nagesh Bhattu Sristy
机构: National Institute of Technology, Andhra Pradesh (安德拉邦国家技术学院)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 9 pages, 3 figures

点击查看摘要

[CV-128] Diverse Text-to-Image Generation via Contrastive Noise Optimization

链接: https://arxiv.org/abs/2510.03813
作者: Byungjun Kim,Soobin Um,Jong Chul Ye
机构: KAIST(韩国科学技术院)
类目: Graphics (cs.GR); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:

点击查看摘要

[CV-129] Road Damage and Manhole Detection using Deep Learning for Smart Cities: A Polygonal Annotation Approach

链接: https://arxiv.org/abs/2510.03797
作者: Rasel Hossen,Diptajoy Mistry,Mushiur Rahman,Waki As Sami Atikur Rahman Hridoy,Sajib Saha,Muhammad Ibrahim
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 13 pages

点击查看摘要

[CV-130] MambaCAFU: Hybrid Multi-Scale and Multi-Attention Model with Mamba-Based Fusion for Medical Image Segmentation

链接: https://arxiv.org/abs/2510.03786
作者: T-Mai Bui,Fares Bougourzi,Fadi Dornaika,Vinh Truong Hoang
机构: University of the Basque Country UPV/EHU, San Sebastian, Spain; IKERBASQUE, Basque Foundation for Science, Bilbao, Spain; Ho Chi Minh City Open University, VietNam
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-131] Efficiency vs. Efficacy: Assessing the Compression Ratio-Dice Score Relationship through a Simple Benchmarking Framework for Cerebrovascular 3D Segmentation

链接: https://arxiv.org/abs/2510.03769
作者: Shimaa Elbana,Ahmad Kamal,Shahd Ahmed Ali,Ahmad Al-Kabbany
机构: Arab Academy for Science and Technology (阿拉伯科技学院); Cairo University (开罗大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Signal Processing (eess.SP)
备注:

点击查看摘要

[CV-132] CoPA: Hierarchical Concept Prompting and Aggregating Network for Explainable Diagnosis MICCAI2025

链接: https://arxiv.org/abs/2510.03767
作者: Yiheng Dong,Yi Lin,Xin Yang
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by MICCAI2025

点击查看摘要

[CV-133] Adaptively Sampling-Reusing-Mixing Decomposed Gradients to Speed Up Sharpness Aware Minimization

链接: https://arxiv.org/abs/2510.03763
作者: Jiaxin Deng,Junbiao Pang
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[CV-134] he Overlooked Value of Test-time Reference Sets in Visual Place Recognition ICCV2025

链接: https://arxiv.org/abs/2510.03751
作者: Mubariz Zaffar,Liangliang Nan,Sebastian Scherer,Julian F. P. Kooij
机构: TU Delft (代尔夫特理工大学); CMU (卡内基梅隆大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at ICCV 2025 Workshop CrocoDL

点击查看摘要

[CV-135] LoRA Patching: Exposing the Frag ility of Proactive Defenses against Deepfakes

链接: https://arxiv.org/abs/2510.03747
作者: Zuomin Qu,Yimao Guo,Qianyue Hu,Wei Lu
机构: Electric Power Research Institute, China Southern Power Grid Company Ltd.(中国南方电网公司电力科学研究院); School of Computer Science and Engineering, Sun Yat-sen University (中山大学计算机科学与工程学院)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-136] Mapping Rio de Janeiros favelas: general-purpose vs. satellite-specific neural networks

链接: https://arxiv.org/abs/2510.03725
作者: Thomas Hallopeau,Joris Guérin,Laurent Demagistri,Youssef Fouzai,Renata Gracie,Vanderlei Pascoal De Matos,Helen Gurgel,Nadine Dessay
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 6 pages, 1 figure, 1 table. Presented at the 21st Brazilian Symposium on Remote Sensing (SBSR 2025)

点击查看摘要

[CV-137] Artery-Vein Segmentation from Fundus Images using Deep Learning

链接: https://arxiv.org/abs/2510.03717
作者: Sharan SK,Subin Sahayam,Umarani Jayaraman,Lakshmi Priya A
机构: IIITDM (印度信息技术研究所); SNU Chennai (斯隆大学钦奈分校)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 12 pages, 6 figures, preprint under review

点击查看摘要

[CV-138] EmbodiSwap for Zero-Shot Robot Imitation Learning

链接: https://arxiv.org/abs/2510.03706
作者: Eadom Dessalene,Pavan Mantripragada,Michael Maynord,Yiannis Aloimonos
机构: University of Maryland, College Park (马里兰大学帕克分校)
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Video link: this https URL

点击查看摘要

[CV-139] Referring Expression Comprehension for Small Objects

链接: https://arxiv.org/abs/2510.03701
作者: Kanoko Goto,Takumi Hirose,Mahiro Ukai,Shuhei Kurita,Nakamasa Inoue
机构: Institute of Science Tokyo (东京科学研究所); National Institute of Informatics (信息基础研究所)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[CV-140] SAMSOD: Rethinking SAM Optimization for RGB-T Salient Object Detection

链接: https://arxiv.org/abs/2510.03689
作者: Zhengyi Liu,Xinrui Wang,Xianyong Fang,Zhengzheng Tu,Linbo Wang
机构: Anhui University (安徽大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by TMM

点击查看摘要

[CV-141] A Novel Cloud-Based Diffusion-Guided Hybrid Model for High-Accuracy Accident Detection in Intelligent Transportation Systems

链接: https://arxiv.org/abs/2510.03675
作者: Siva Sai,Saksham Gupta,Vinay Chamola,Rajkumar Buyya
机构: BITS-Pilani, Pilani Campus (BITS-比拉尼学院); The University of Melbourne (墨尔本大学); qCLOUDS Laboratory (量子云计算与分布式系统实验室)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-142] MonitorVLM:A Vision Language Framework for Safety Violation Detection in Mining Operations

链接: https://arxiv.org/abs/2510.03666
作者: Jiang Wu,Sichao Wu,Yinsong Ma,Guangyuan Yu,Haoyuan Xu,Lifang Zheng,Jingliang Duan
机构: University of Science and Technology Beijing (北京科技大学); Johns Hopkins University (约翰霍普金斯大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[CV-143] Diffusion-Classifier Synergy: Reward-Aligned Learning via Mutual Boosting Loop for FSCIL NEURIPS2025

链接: https://arxiv.org/abs/2510.03608
作者: Ruitao Wu,Yifan Zhao,Guangyao Chen,Jia Li
机构: Beihang University (北京航空航天大学); Zhongguancun Academy; Peking University (北京大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by NeurIPS 2025

点击查看摘要

[CV-144] Unsupervised Transformer Pre-Training for Images: Self-Distillation Mean Teachers and Random Crops

链接: https://arxiv.org/abs/2510.03606
作者: Mattia Scardecchia
机构: Technical University Munich (慕尼黑工业大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Image and Video Processing (eess.IV)
备注:

点击查看摘要

[CV-145] Exploring the Hierarchical Reasoning Model for Small Natural-Image Classification Without Augmentation

链接: https://arxiv.org/abs/2510.03598
作者: Alexander V. Mantzaris
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:

点击查看摘要

[CV-146] A Hybrid Co-Finetuning Approach for Visual Bug Detection in Video Games AAAI

链接: https://arxiv.org/abs/2510.03591
作者: Faliu Yi,Sherif Abdelfattah,Wei Huang,Adrian Brown
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted at the 21st AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment (AIIDE 2025)

点击查看摘要

[CV-147] FrameOracle: Learning What to See and How Much to See in Videos

链接: https://arxiv.org/abs/2510.03584
作者: Chaoyu Li,Tianzhi Li,Fei Tao,Zhenyu Zhao,Ziqian Wu,Maozheng Zhao,Juntong Song,Cheng Niu,Pooyan Fazli
机构: NewsBreak; Arizona State University (亚利桑那州立大学); Carnegie Mellon University (卡内基梅隆大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-148] Efficient Test-Time Scaling for Small Vision-Language Models

链接: https://arxiv.org/abs/2510.03574
作者: Mehmet Onurcan Kaya,Desmond Elliott,Dim P. Papadopoulos
机构: 未知
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-149] Evaluating OCR performance on food packaging labels in South Africa

链接: https://arxiv.org/abs/2510.03570
作者: Mayimunah Nagayi,Alice Khan,Tamryn Frank,Rina Swart,Clement Nyirenda
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 17 pages

点击查看摘要

[CV-150] Longitudinal Flow Matching for Trajectory Modeling

链接: https://arxiv.org/abs/2510.03569
作者: Mohammad Mohaiminul Islam,Thijs P. Kuipers,Sharvaree Vadgama,Coen de Vente,Afsana Khan,Clara I. Sánchez,Erik J. Bekkers
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (stat.ML)
备注:

点击查看摘要

[CV-151] Real-Time Assessment of Bystander Situation Awareness in Drone-Assisted First Aid

链接: https://arxiv.org/abs/2510.03558
作者: Shen Chang,Renran Tian,Nicole Adams,Nan Kong
机构: 1. University of Texas at Dallas (德克萨斯大学达拉斯分校); 2. Google (谷歌); 3. Microsoft (微软); 4. Southern Methodist University (南卫理公会大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-152] GAS-MIL: Group-Aggregative Selection Multi-Instance Learning for Ensemble of Foundation Models in Digital Pathology Image Analysis

链接: https://arxiv.org/abs/2510.03555
作者: Peiran Quan,Zifan Gu,Zhuo Zhao,Qin Zhou,Donghan M. Yang,Ruichen Rong,Yang Xie,Guanghua Xiao
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[CV-153] Streaming Drag -Oriented Interactive Video Manipulation: Drag Anything Anytime!

链接: https://arxiv.org/abs/2510.03550
作者: Junbao Zhou,Yuan Zhou,Kesen Zhao,Qingshan Xu,Beier Zhu,Richang Hong,Hanwang Zhang
机构: Nanyang Technological University (南洋理工大学); Hefei University of Technology (合肥工业大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-154] Unmasking Puppeteers: Leverag ing Biometric Leakage to Disarm Impersonation in AI-based Videoconferencing

【速读】：该论文旨在解决基于生成式AI（Generative AI）的虚拟头像视频会议系统中存在的身份伪造安全问题，即攻击者可通过操控传输的姿势-表情潜在表示（pose-expression latent）劫持受害者的人脸形象进行实时欺骗，而传统深度伪造检测方法因视频全为合成内容而失效。解决方案的关键在于利用一个核心观察：该潜在表示中隐含了驱动身份的生物特征信息。作者提出了一种无需查看重建RGB视频的生物特征泄露防御机制——一种基于姿态条件的大 margin 对比编码器，能够从潜在表示中解耦出稳定的身份线索并抑制瞬时的姿态和表情干扰；通过在该解耦嵌入上执行简单的余弦相似度测试，即可在视频渲染过程中实时识别非法身份替换行为。

链接: https://arxiv.org/abs/2510.03548
作者: Danial Samadi Vahdati,Tai Duc Nguyen,Ekta Prashnani,Koki Nagano,David Luebke,Orazio Gallo,Matthew Stamm
机构: Drexel University (德雷塞尔大学); NVIDIA Corporation (英伟达公司)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

Abstract:AI-based talking-head videoconferencing systems reduce bandwidth by sending a compact pose-expression latent and re-synthesizing RGB at the receiver, but this latent can be puppeteered, letting an attacker hijack a victim’s likeness in real time. Because every frame is synthetic, deepfake and synthetic video detectors fail outright. To address this security problem, we exploit a key observation: the pose-expression latent inherently contains biometric information of the driving identity. Therefore, we introduce the first biometric leakage defense without ever looking at the reconstructed RGB video: a pose-conditioned, large-margin contrastive encoder that isolates persistent identity cues inside the transmitted latent while cancelling transient pose and expression. A simple cosine test on this disentangled embedding flags illicit identity swaps as the video is rendered. Our experiments on multiple talking-head generation models show that our method consistently outperforms existing puppeteering defenses, operates in real-time, and shows strong generalization to out-of-distribution scenarios.
zh

[CV-155] SketchPlan: Diffusion Based Drone Planning From Human Sketches

链接: https://arxiv.org/abs/2510.03545
作者: Sixten Norelius,Aaron O. Feldman,Mac Schwager
机构: KTH Royal Institute of Technology (皇家理工学院); Stanford University (斯坦福大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: Code available at this https URL

点击查看摘要

[CV-156] From Scope to Script: An Automated Report Generation Model for Gastrointestinal Endoscopy

链接: https://arxiv.org/abs/2510.03543
作者: Evandros Kaklamanos,Kristjana Kristinsdottir,Jonathan Huang,Dustin Carlson,Rajesh Keswani,John Pandolfino,Mozziyar Etemadi
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-157] Domain Generalization for Semantic Segmentation: A Survey CVPR2025

链接: https://arxiv.org/abs/2510.03540
作者: Manuel Schwonberg,Hanno Gottschalk
机构: TU Berlin (柏林工业大学); CARIAD SE
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to CVPR2025W

点击查看摘要

[CV-158] Efficient Surgical Robotic Instrument Pose Reconstruction in Real World Conditions Using Unified Feature Detection

链接: https://arxiv.org/abs/2510.03532
作者: Zekai Liang,Kazuya Miyata,Xiao Liang,Florian Richter,Michael C. Yip
机构: University of California San Diego (加州大学圣地亚哥分校)
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-159] Platonic Transformers: A Solid Choice For Equivariance

链接: https://arxiv.org/abs/2510.03511
作者: Mohammad Mohaiminul Islam,Rishabh Anand,David R. Wessels,Friso de Kruiff,Thijs P. Kuipers,Rex Ying,Clara I. Sánchez,Sharvaree Vadgama,Georg Bökman,Erik J. Bekkers
机构: QurAI(量子人工智能实验室); Univ. of Amsterdam(阿姆斯特丹大学); Yale University(耶鲁大学); BMEP(生物医学工程与物理中心); Amsterdam UMC(阿姆斯特丹大学医疗中心); AMLab(阿姆斯特丹机器学习实验室)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Image and Video Processing (eess.IV)
备注:

点击查看摘要

[CV-160] Real-Time Threaded Houbara Detection and Segmentation for Wildlife Conservation using Mobile Platforms

链接: https://arxiv.org/abs/2510.03501
作者: Lyes Saad Saoud,Loic Lesobre,Enrico Sorato,Irfan Hussain
机构: Kuwait University (科威特大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:

点击查看摘要

[CV-161] DuPLUS: Dual-Prompt Vision-Language Framework for Universal Medical Image Segmentation and Prognosis

链接: https://arxiv.org/abs/2510.03483
作者: Numan Saeed,Tausifa Jan Saleem,Fadillah Maani,Muhammad Ridzuan,Hu Wang,Mohammad Yaqub
机构: Mohamed bin Zayed University of Artificial Intelligence (MBZUAI)(穆罕默德·本·扎耶德人工智能大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[CV-162] PEaRL: Pathway-Enhanced Representation Learning for Gene and Pathway Expression Prediction from Histology

链接: https://arxiv.org/abs/2510.03455
作者: Sejuti Majumder,Saarthak Kapse,Moinak Bhattacharya,Xuan Xu,Alisa Yurovsky,Prateek Prasanna
机构: Stony Brook University (石溪大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-163] Denoising of Two-Phase Optically Sectioned Structured Illumination Reconstructions Using Encoder-Decoder Networks ICASSP2026

链接: https://arxiv.org/abs/2510.03452
作者: Allison Davis,Yezhi Shen,Xiaoyu Ji,Fengqing Zhu
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 4 figures, submitted to ICASSP 2026

点击查看摘要

[CV-164] Spatial-ViLT: Enhancing Visual Spatial Reasoning through Multi-Task Learning

链接: https://arxiv.org/abs/2510.03441
作者: Chashi Mahiul Islam,Oteo Mamo,Samuel Jacob Chacko,Xiuwen Liu,Weikuan Yu
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 12 pages, 5 figures

点击查看摘要

[CV-165] Visual Language Model as a Judge for Object Detection in Industrial Diagrams ICASSP2026

链接: https://arxiv.org/abs/2510.03376
作者: Sanjukta Ghosh
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
备注: Pre-review version submitted to IEEE ICASSP 2026

点击查看摘要

[CV-166] Conditional Pseudo-Supervised Contrast for Data-Free Knowledge Distillation

链接: https://arxiv.org/abs/2510.03375
作者: Renrong Shao,Wei Zhang,Jun wang
机构: 未知
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 13 pages

点击查看摘要

[CV-167] Unified Unsupervised Anomaly Detection via Matching Cost Filtering

链接: https://arxiv.org/abs/2510.03363
作者: Zhe Zhang,Mingxiu Cai,Gaochang Wu,Jing Zhang,Lingqiao Liu,Dacheng Tao,Tianyou Chai,Xiatian Zhu
机构: State Key Laboratory of Synthetical Automation for Process Industries, Northeastern University, Shenyang, China; School of Computer Science, Wuhan University, China; School of Computer Science, The University of Adelaide, Australia; College of Computing & Data Science, Nanyang Technological University, Singapore; Surrey Institute for People-Centred Artificial Intelligence, and Centre for Vision, Speech and Signal Processing, University of Surrey, Guildford, UK
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Image and Video Processing (eess.IV)
备注: 63 pages (main paper and supplementary material), 39 figures, 58 tables. Submitted to IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)

点击查看摘要

[CV-168] Provenance Networks: End-to-End Exemplar-Based Explainability

链接: https://arxiv.org/abs/2510.03361
作者: Ali Kayyam,Anusha Madan Gopal,M. Anthony Lewis
机构: BrainChip Inc.(BrainChip公司)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[CV-169] Learned Display Radiance Fields with Lensless Cameras

链接: https://arxiv.org/abs/2510.03356
作者: Ziyang Chen,Yuta Itoh,Kaan Akşit
机构: University College London (伦敦大学学院); University of Tokyo (东京大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Emerging Technologies (cs.ET)
备注:

点击查看摘要

[CV-170] Sonar Image Datasets: A Comprehensive Survey of Resources Challenges and Applications

链接: https://arxiv.org/abs/2510.03353
作者: Larissa S. Gomes,Gustavo P. Almeida,Bryan U. Moreira,Marco Quiroz,Breno Xavier,Lucas Soares,Stephanie L. Brião,Felipe G. Oliveira,Paulo L. J. Drews-Jr
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Published in the Conference on Graphics, Patterns and Images (SIBGRAPI). This 4-page paper presents a timeline of publicly available datasets up to the year 2025

点击查看摘要

[CV-171] Inference-Time Search using Side Information for Diffusion-based Image Reconstruction

链接: https://arxiv.org/abs/2510.03352
作者: Mahdi Farahbakhsh,Vishnu Teja Kunde,Dileep Kalathil,Krishna Narayanan,Jean-Francois Chamberland
机构: Texas A&M University (德州农工大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[CV-172] Visual Odometry with Transformers

链接: https://arxiv.org/abs/2510.03348
作者: Vlardimir Yugay,Duy-Kien Nguyen,Theo Gevers,Cees G. M. Snoek,Martin R. Oswald
机构: University of Amsterdam (阿姆斯特丹大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-173] OpusAnimation: Code-Based Dynamic Chart Generation

链接: https://arxiv.org/abs/2510.03341
作者: Bozheng Li,Miao Yang,Zhenhan Chen,Jiawang Cao,Mushui Liu,Yi Lu,Yongliang Wu,Bin Zhang,Yangguang Ji,Licheng Tang,Jay Wu,Wenbo Zhu
机构: Opus AI Research; Brown University (布朗大学); Zhejiang University (浙江大学); University of Toronto (多伦多大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: working in progress

点击查看摘要

[CV-174] Error correction in multiclass image classification of facial emotion on unbalanced samples

链接: https://arxiv.org/abs/2510.03337
作者: Andrey A. Lebedev,Victor B. Kazantsev,Sergey V. Stasenko
机构: Lobachevsky State University of Nizhny Novgorod (下诺夫哥罗德国立大学); Moscow Center for Advanced Studies (莫斯科高级研究所)
类目: Computer Vision and Pattern Recognition (cs.CV); Neurons and Cognition (q-bio.NC)
备注:

点击查看摘要

[CV-175] DECOR: Deep Embedding Clustering with Orientation Robustness

链接: https://arxiv.org/abs/2510.03328
作者: Fiona Victoria Stanley Jothiraj,Arunaggiri Pandian Karunanidhi,Seth A. Eichmeyer
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:

点击查看摘要

[CV-176] Advances in Medical Image Segmentation: A Comprehensive Survey with a Focus on Lumbar Spine Applications

链接: https://arxiv.org/abs/2510.03318
作者: Ahmed Kabil,Ghada Khoriba,Mina Yousef,Essam A. Rashed
机构: New Valley University (新谷大学); University of Hyogo (兵库县立大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Computers in Biology and Medicine (to appear)

点击查看摘要

[CV-177] Photorealistic Inpainting for Perturbation-based Explanations in Ecological Monitoring NEURIPS2025

链接: https://arxiv.org/abs/2510.03317
作者: Günel Aghakishiyeva,Jiayi Zhou,Saagar Arya,James David Poling,Holly R. Houliston,Jamie N. Womble,David W. Johnston,Brinnae Bent
机构: Duke University (杜克大学); University of Agder (奥德大学); University of Cambridge (剑桥大学); U.S. National Park Service (美国国家公园管理局)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted to NeurIPS 2025 Imageomics Workshop

点击查看摘要

[CV-178] he View From Space: Navigating Instrumentation Differences with EOFMs

链接: https://arxiv.org/abs/2510.03316
作者: Ryan P. Demilt,Nicholas LaHaye,Karis Tenneson
机构: Spatial Informatics Group (空间信息集团)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[CV-179] A Comprehensive Review on Artificial Intelligence Empowered Solutions for Enhancing Pedestrian and Cyclist Safety

链接: https://arxiv.org/abs/2510.03314
作者: Shucheng Zhang,Yan Shi,Bingzhang Wang,Yuang Zhang,Muhammad Monjurul Karim,Kehua Chen,Chenxi Liu,Mehrdad Nasri,Yinhai Wang
机构: University of Washington (华盛顿大学); University of Utah (犹他大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 20 pages, 4 figures, 5 tables

点击查看摘要

[CV-180] Universal Beta Splatting

链接: https://arxiv.org/abs/2510.03312
作者: Rong Liu,Zhongpai Gao,Benjamin Planche,Meida Chen,Van Nguyen Nguyen,Meng Zheng,Anwesa Choudhuri,Terrence Chen,Yue Wang,Andrew Feng,Ziyan Wu
机构: University of Southern California (南加州大学); United Imaging Intelligence (联影智能)
类目: Graphics (cs.GR); Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
备注:

点击查看摘要

[CV-181] Creative synthesis of kinematic mechanisms

链接: https://arxiv.org/abs/2510.03308
作者: Jiong Lin,Jialong Ning,Judah Goldfeder,Hod Lipson
机构: Creative Machines Lab, Columbia University (哥伦比亚大学)
类目: Graphics (cs.GR); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 6pages, 6 figures

点击查看摘要

[CV-182] Revoking Amnesia: RL-based Trajectory Optimization to Resurrect Erased Concepts in Diffusion Models

链接: https://arxiv.org/abs/2510.03302
作者: Daiheng Gao,Nanxiang Jiang,Andi Zhang,Shilin Lu,Yufei Tang,Wenbo Zhou,Weiming Zhang,Zhaoxin Fan
机构: USTC; Beihang University; University of Manchester; NTU; FYUST
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 21 pages, 10 figures

点击查看摘要

[CV-183] Convolutional Neural Nets vs Vision Transformers: A SpaceNet Case Study with Balanced vs Imbalanced Regimes

链接: https://arxiv.org/abs/2510.03297
作者: Akshar Gothi
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 5 pages, 1 figure, 9 tables. Code and artifacts: this https URL (release v1.0.1)

点击查看摘要

[CV-184] Domain-Robust Marine Plastic Detection Using Vision Models

链接: https://arxiv.org/abs/2510.03294
作者: Saanvi Kataria
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 16 pages, 5 figures, 1 table

点击查看摘要

[CV-185] Visualizing Celebrity Dynamics in Video Content: A Proposed Approach Using Face Recognition Timestamp Data

链接: https://arxiv.org/abs/2510.03292
作者: Doğanay Demir,İlknur Durgar Elkahlout
机构: TRT(土耳其广播电视台)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-186] SoC-DT: Standard-of-Care Aligned Digital Twins for Patient-Specific Tumor Dynamics

链接: https://arxiv.org/abs/2510.03287
作者: Moinak Bhattacharya,Gagandeep Singh,Prateek Prasanna
机构: Stony Brook University (石溪大学); Columbia University (哥伦比亚大学)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-187] SDQ-LLM : Sigma-Delta Quantization for 1-bit LLM s of any size

链接: https://arxiv.org/abs/2510.03275
作者: Junhao Xia,Ming Zhao,Limin Xiao,Xiujun Zhang
机构: Tsinghua University (清华大学); Beijing National Research Center for Information Science and Technology (北京信息科学与技术国家研究中心)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-188] Rethinking Inter-LoRA Orthogonality in Adapter Merging: Insights from Orthogonal Monte Carlo Dropout

链接: https://arxiv.org/abs/2510.03262
作者: Andi Zhang,Xuan Ding,Haofan Wang,Steven McDonagh,Samuel Kaski
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-189] Universal Multi-Domain Translation via Diffusion Routers

链接: https://arxiv.org/abs/2510.03252
作者: Duc Kieu,Kien Do,Tuan Hoang,Thao Minh Le,Tung Kieu,Dang Nguyen,Thin Nguyen
机构: Deakin University (迪肯大学); Pennsylvania State University (宾夕法尼亚州立大学); Aalborg University (奥尔堡大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-190] Real-Time Brain Biomechanics Prediction with Neural Operators: Toward Clinically Deployable Traumatic Brain Injury Models

链接: https://arxiv.org/abs/2510.03248
作者: Anusha Agarwal,Dibakar Roy Sarkar,Somdatta Goswami
机构: Johns Hopkins Whiting School of Engineering (约翰霍普金斯大学怀廷工程学院)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Medical Physics (physics.med-ph)
备注:

点击查看摘要

[CV-191] Frequency-Aware Model Parameter Explorer: A new attribution method for improving explainability

链接: https://arxiv.org/abs/2510.03245
作者: Ali Yavari,Alireza Mohamadi,Elham Beydaghi,Rainer A. Leitgeb
机构: Medical University of Vienna (维也纳医科大学); Sharif University of Technology (沙里夫理工大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint

点击查看摘要

[CV-192] VIFO: Visual Feature Empowered Multivariate Time Series Forecasting with Cross-Modal Fusion

链接: https://arxiv.org/abs/2510.03244
作者: Yanlong Wang,Hang Yu,Jian Xu,Fei Ma,Hongkang Zhang,Tongtong Feng,Zijian Zhang,Shao-Lun Huang,Danny Dongning Sun,Xiao-Ping Zhang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-193] xtured Gaussians for Enhanced 3D Scene Appearance Modeling CVPR2025

链接: https://arxiv.org/abs/2411.18625
作者: Brian Chao,Hung-Yu Tseng,Lorenzo Porzi,Chen Gao,Tuotuo Li,Qinbo Li,Ayush Saraf,Jia-Bin Huang,Johannes Kopf,Gordon Wetzstein,Changil Kim
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Graphics (cs.GR); Image and Video Processing (eess.IV)
备注: Will be presented at CVPR 2025. Project website: this https URL

点击查看摘要

[CV-194] A Modular Conditional Diffusion Framework for Image Reconstruction

链接: https://arxiv.org/abs/2411.05993
作者: Magauiya Zhussip,Iaroslav Koshelev,Stamatis Lefkimmiatis
机构: MTS AI; AI Foundation and Algorithm Lab
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Image and Video Processing (eess.IV)
备注:

点击查看摘要

[CV-195] Adaptive double-phase Rudin–Osher–Fatemi denoising model

链接: https://arxiv.org/abs/2510.04382
作者: Wojciech Górny,Michał Łasica,Alexandros Matsoukas
机构: University of Vienna (维也纳大学); University of Warsaw (华沙大学); Polish Academy of Sciences (波兰科学院); National Technical University of Athens (雅典国立技术大学)
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV); Numerical Analysis (math.NA)
备注: 21 pages, 18 figures, supplementary material available at: this https URL

点击查看摘要

[CV-196] he method of the approximate inverse for limited-angle CT

链接: https://arxiv.org/abs/2510.04369
作者: Bernadette Hahn,Gael Rigaud,Richard Schmähl
机构: 未知
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV); Numerical Analysis (math.NA)
备注:

点击查看摘要

[CV-197] MoME: Mixture of Matryoshka Experts for Audio-Visual Speech Recognition NEURIPS2025

链接: https://arxiv.org/abs/2510.04136
作者: Umberto Cappellazzo,Minsu Kim,Pingchuan Ma,Honglie Chen,Xubo Liu,Stavros Petridis,Maja Pantic
机构: Imperial College London (帝国理工学院); Meta AI (Meta人工智能实验室); NatWest AI Research (英国国民西敏寺银行人工智能研究)
类目: Audio and Speech Processing (eess.AS); Computer Vision and Pattern Recognition (cs.CV); Sound (cs.SD)
备注: NeurIPS 2025

点击查看摘要

[CV-198] Super-resolution image projection over an extended depth of field using a diffractive decoder

链接: https://arxiv.org/abs/2510.03938
作者: Hanlong Chen,Cagatay Isil,Tianyi Gan,Mona Jarrahi,Aydogan Ozcan
机构: 未知
类目: Optics (physics.optics); Computer Vision and Pattern Recognition (cs.CV); Neural and Evolutionary Computing (cs.NE); Applied Physics (physics.app-ph)
备注: 18 Pages, 6 Figures

点击查看摘要

[CV-199] Sliding Window Attention for Learned Video Compression

链接: https://arxiv.org/abs/2510.03926
作者: Alexander Kopte,André Kaup
机构: 未知
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted for PCS 2025

点击查看摘要

[CV-200] AI-Assisted Pleural Effusion Volume Estimation from Contrast-Enhanced CT Images

链接: https://arxiv.org/abs/2510.03856
作者: Sanhita Basu,Tomas Fröding,Ali Teymur Kahraman,Dimitris Toumpanakis,Tobias Sjöblom
机构: 未知
类目: Image and Video Processing (eess.IV); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-201] owards Robust and Generalizable Continuous Space-Time Video Super-Resolution with Events

链接: https://arxiv.org/abs/2510.03833
作者: Shuoyan Wei,Feng Li,Shengeng Tang,Runmin Cong,Yao Zhao,Meng Wang,Huihui Bai
机构: Beijing Jiaotong University (北京交通大学); Hefei University of Technology (合肥工业大学); Shandong University (山东大学)
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
备注: 17 pages, 12 figures, 14 tables. Under review

点击查看摘要

[CV-202] Model-Guided Microstimulation Steers Primate Visual Behavior

链接: https://arxiv.org/abs/2510.03684
作者: Johannes Mehrer,Ben Lonnqvist,Anna Mitola,Abdulkadir Gokce,Paolo Papale,Martin Schrimpf
机构: EPFL(瑞士联邦理工学院); Netherlands Institute for Neuroscience(荷兰神经科学研究所); University of Parma(帕尔马大学)
类目: Neurons and Cognition (q-bio.NC); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

[CV-203] How We Won BraTS-SSA 2025: Brain Tumor Segmentation in the Sub-Saharan African Population Using Segmentation-Aware Data Augmentation and Model Ensembling MICCAI

链接: https://arxiv.org/abs/2510.03568
作者: Claudia Takyi Ankomah,Livingstone Eli Ayivor,Ireneaus Nyame,Leslie Wambo,Patrick Yeboah Bonsu,Aondona Moses Iorumbur,Raymond Confidence,Toufiq Musah
机构: 未知
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注: Brain Tumor Segmentation Challenge, Medical Image Computing and Computer Assisted Intervention (MICCAI) Conference, 11 Pages, 2 Figures, 2 Tables

点击查看摘要

[CV-204] Real-time nonlinear inversion of magnetic resonance elastography with operator learning

链接: https://arxiv.org/abs/2510.03372
作者: Juampablo E. Heras Rivera,Caitlin M. Neher,Mehmet Kurt
机构: 未知
类目: Image and Video Processing (eess.IV); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:

点击查看摘要

人工智能

[AI-0] opInG: Topologically Interpretable Graph Learning via Persistent Rationale Filtration ICML2025

【速读】：该论文旨在解决图神经网络（Graph Neural Networks, GNNs）在关键决策场景中因缺乏可解释性而难以被采纳的问题。现有方法在面对复杂且多样的理由子图（rationale subgraphs）时表现不佳，难以稳定识别出具有持久拓扑结构的解释依据。其解决方案的关键在于提出TopInG：一种基于持久同调（persistent homology）的拓扑可解释图学习框架，通过理由过滤学习（rationale filtration learning）建模理由子图的自回归生成过程，并引入自适应拓扑约束——拓扑差异（topological discrepancy），以确保理由子图与无关子图之间保持稳定的拓扑区分度。该方法在理论上保证了损失函数在特定条件下唯一优化于真实理由子图，实验证明其能有效提升预测准确率与解释质量，同时缓解虚假相关性问题。

链接: https://arxiv.org/abs/2510.05102
作者: Cheng Xin,Fan Xu,Xin Ding,Jie Gao,Jiaxin Ding
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computational Geometry (cs.CG); Algebraic Topology (math.AT); Machine Learning (stat.ML)
备注: submitted to ICML 2025

点击查看摘要

Abstract:Graph Neural Networks (GNNs) have shown remarkable success across various scientific fields, yet their adoption in critical decision-making is often hindered by a lack of interpretability. Recently, intrinsically interpretable GNNs have been studied to provide insights into model predictions by identifying rationale substructures in graphs. However, existing methods face challenges when the underlying rationale subgraphs are complex and varied. In this work, we propose TopInG: Topologically Interpretable Graph Learning, a novel topological framework that leverages persistent homology to identify persistent rationale subgraphs. TopInG employs a rationale filtration learning approach to model an autoregressive generation process of rationale subgraphs, and introduces a self-adjusted topological constraint, termed topological discrepancy, to enforce a persistent topological distinction between rationale subgraphs and irrelevant counterparts. We provide theoretical guarantees that our loss function is uniquely optimized by the ground truth under specific conditions. Extensive experiments demonstrate TopInG’s effectiveness in tackling key challenges, such as handling variform rationale subgraphs, balancing predictive performance with interpretability, and mitigating spurious correlations. Results show that our approach improves upon state-of-the-art methods on both predictive accuracy and interpretation quality.
zh

[AI-1] Staircase Streaming for Low-Latency Multi-Agent Inference

链接: https://arxiv.org/abs/2510.05059
作者: Junlin Wang,Jue Wang,Zhen(Zach)Xu,Ben Athiwaratkun,Bhuwan Dhingra,Ce Zhang,James Zou
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-2] HybridFlow: Quantification of Aleatoric and Epistemic Uncertainty with a Single Hybrid Model

链接: https://arxiv.org/abs/2510.05054
作者: Peter Van Katwyk,Karianne J. Bergen
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Reviewed and published in TMLR at this https URL

点击查看摘要

[AI-3] Look-ahead Reasoning with a Learned Model in Imperfect Information Games

链接: https://arxiv.org/abs/2510.05048
作者: Ondřej Kubíček,Viliam Lisý
机构: 未知
类目: Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT)
备注:

点击查看摘要

[AI-4] st-Time Scaling in Diffusion LLM s via Hidden Semi-Autoregressive Experts

【速读】：该论文旨在解决扩散型大语言模型（diffusion-based large language models, dLLMs）在推理阶段如何有效利用训练时学到的极端依赖关系的问题。现有方法通常采用单一固定推理调度策略，导致无法充分利用模型隐含的多样化生成行为。解决方案的关键在于提出HEX（Hidden semi-autoregressive EXperts for test-time scaling），一种无需额外训练的推理方法，通过在不同块大小的生成路径上进行多数投票，实现对异构块调度的集成，从而鲁棒地规避单一调度带来的失败模式，显著提升多个推理基准上的性能表现。

链接: https://arxiv.org/abs/2510.05040
作者: Jihoon Lee,Hoyeon Moon,Kevin Zhai,Arun Kumar Chithanar,Anit Kumar Sahu,Soummya Kar,Chul Lee,Souradip Chakraborty,Amrit Singh Bedi
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

Abstract:Diffusion-based large language models (dLLMs) are trained flexibly to model extreme dependence in the data distribution; however, how to best utilize this information at inference time remains an open problem. In this work, we uncover an interesting property of these models: dLLMs trained on textual data implicitly learn a mixture of semi-autoregressive experts, where different generation orders reveal different specialized behaviors. We show that committing to any single, fixed inference time schedule, a common practice, collapses performance by failing to leverage this latent ensemble. To address this, we introduce HEX (Hidden semiautoregressive EXperts for test-time scaling), a training-free inference method that ensembles across heterogeneous block schedules. By doing a majority vote over diverse block-sized generation paths, HEX robustly avoids failure modes associated with any single fixed schedule. On reasoning benchmarks such as GSM8K, it boosts accuracy by up to 3.56X (from 24.72% to 88.10%), outperforming top-K margin inference and specialized fine-tuned methods like GRPO, without additional training. HEX even yields significant gains on MATH benchmark from 16.40% to 40.00%, scientific reasoning on ARC-C from 54.18% to 87.80%, and TruthfulQA from 28.36% to 57.46%. Our results establish a new paradigm for test-time scaling in diffusion-based LLMs (dLLMs), revealing that the sequence in which masking is performed plays a critical role in determining performance during inference.
zh

[AI-5] Graph-Aware Diffusion for Signal Generation

链接: https://arxiv.org/abs/2510.05036
作者: Sergio Rozada,Vimal K. B.,Andrea Cavallo,Antonio G. Marques,Hadi Jamali-Rad,Elvin Isufi
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-6] Rethinking Langevin Thompson Sampling from A Stochastic Approximation Perspective

链接: https://arxiv.org/abs/2510.05023
作者: Weixin Wang,Haoyang Zheng,Guang Lin,Wei Deng,Pan Xu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注: 39 pages, 3 figures, 2 tables

点击查看摘要

[AI-7] hink Then Embed: Generative Context Improves Multimodal Embedding

链接: https://arxiv.org/abs/2510.05014
作者: Xuanming Cui,Jianpeng Cheng,Hong-you Chen,Satya Narayan Shukla,Abhijeet Awasthi,Xichen Pan,Chaitanya Ahuja,Shlok Kumar Mishra,Qi Guo,Ser-Nam Lim,Aashu Singh,Xiangjun Fan
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[AI-8] AutoEmpirical: LLM -Based Automated Research for Empirical Software Fault Analysis

链接: https://arxiv.org/abs/2510.04997
作者: Jiongchi Yu,Weipeng Jiang,Xiaoyu Zhang,Qiang Hu,Xiaofei Xie,Chao Shen
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 5 pages

点击查看摘要

[AI-9] Aligning Perception Reasoning Modeling and Interaction: A Survey on Physical AI

链接: https://arxiv.org/abs/2510.04978
作者: Kun Xiang,Terry Jingchen Zhang,Yinya Huang,Jixi He,Zirong Liu,Yueling Tang,Ruizhe Zhou,Lijing Luo,Youpeng Wen,Xiuwei Chen,Bingqian Lin,Jianhua Han,Hang Xu,Hanhui Li,Bin Dong,Xiaodan Liang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-10] Safe and Compliant Cross-Market Trade Execution via Constrained RL and Zero-Knowledge Audits

链接: https://arxiv.org/abs/2510.04952
作者: Ailiya Borjigin,Cong He
机构: 未知
类目: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC)
备注: 22 pages, 2 figures

点击查看摘要

[AI-11] Feasibility-Aware Decision-Focused Learning for Predicting Parameters in the Constraints

链接: https://arxiv.org/abs/2510.04951
作者: Jayanta Mandi,Marianne Defresne,Senne Berden,Tias Guns
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-12] Federated Self-Supervised Learning for Automatic Modulation Classification under Non-IID and Class-Imbalanced Data

链接: https://arxiv.org/abs/2510.04927
作者: Usman Akram,Yiyue Chen,Haris Vikalo
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Signal Processing (eess.SP)
备注:

点击查看摘要

[AI-13] Glocal Information Bottleneck for Time Series Imputation

【速读】：该论文针对时间序列插补（Time Series Imputation, TSI）中高缺失率下模型性能下降的问题展开研究，指出现有方法通常仅优化点对点重建损失（point-wise reconstruction loss），虽在训练阶段表现良好，但在推理阶段却因缺乏全局信息引导而产生不良插补结果和失真的潜在表示分布（latent representation distributions），从而导致过拟合局部噪声、无法捕捉数据全局结构。解决方案的关键在于提出一种新的训练范式——Glocal Information Bottleneck（Glocal-IB），其核心创新是扩展标准信息瓶颈（Information Bottleneck, IB）框架，引入一个基于可计算互信息近似的全局对齐损失（Global Alignment loss），该损失强制掩码输入的潜在表示与原始观测对应样本的潜在表示对齐，从而在保留局部细节的同时抑制缺失值带来的噪声干扰，提升模型在高缺失率下的泛化能力。

链接: https://arxiv.org/abs/2510.04910
作者: Jie Yang,Kexin Zhang,Guibin Zhang,Philip S. Yu,Kaize Ding
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

Abstract:Time Series Imputation (TSI), which aims to recover missing values in temporal data, remains a fundamental challenge due to the complex and often high-rate missingness in real-world scenarios. Existing models typically optimize the point-wise reconstruction loss, focusing on recovering numerical values (local information). However, we observe that under high missing rates, these models still perform well in the training phase yet produce poor imputations and distorted latent representation distributions (global information) in the inference phase. This reveals a critical optimization dilemma: current objectives lack global guidance, leading models to overfit local noise and fail to capture global information of the data. To address this issue, we propose a new training paradigm, Glocal Information Bottleneck (Glocal-IB). Glocal-IB is model-agnostic and extends the standard IB framework by introducing a Global Alignment loss, derived from a tractable mutual information approximation. This loss aligns the latent representations of masked inputs with those of their originally observed counterparts. It helps the model retain global structure and local details while suppressing noise caused by missing values, giving rise to better generalization under high missingness. Extensive experiments on nine datasets confirm that Glocal-IB leads to consistently improved performance and aligned latent representations under missingness. Our code implementation is available in this https URL.
zh

[AI-14] Focused Skill Discovery: Learning to Control Specific State Variables while Minimizing Side Effects

链接: https://arxiv.org/abs/2510.04901
作者: Jonathan Colaço Carr,Qinyi Sun,Cameron Allen
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Reinforcement Learning Journal 2025

点击查看摘要

[AI-15] Human Behavior Atlas: Benchmarking Unified Psychological and Social Behavior Understanding

【速读】：该论文旨在解决如何通过智能系统感知心理与社会行为（psychological and social behaviors）这一挑战，其核心难点在于这些行为所反映的内在情感状态（affective states）、认知状态（cognitive states）以及病理状态（pathologies）具有高度复杂性、多维性和个体差异性。现有研究多依赖于专用数据集和单任务模型，难以实现可扩展性、跨任务迁移和广泛泛化能力。解决方案的关键在于构建一个统一的基准——Human Behavior Atlas，该基准整合了超过10万条文本、音频和视觉模态样本，涵盖情感、认知、病理及社交过程等多样化行为任务，从而减少冗余、提升训练效率，并增强行为特征在不同领域间的泛化能力。基于此基准，作者训练了三种模型（OmniSapiens-7B SFT、BAM 和 RL），实验证明其在多种行为任务中均优于现有多模态大语言模型（multimodal LLMs），且预训练后对新行为数据集具有显著迁移性能提升。

链接: https://arxiv.org/abs/2510.04899
作者: Keane Ong,Wei Dai,Carol Li,Dewei Feng,Hengzhi Li,Jingyao Wu,Jiaee Cheong,Rui Mao,Gianmarco Mengaldo,Erik Cambria,Paul Pu Liang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

Abstract:Using intelligent systems to perceive psychological and social behaviors, that is, the underlying affective, cognitive, and pathological states that are manifested through observable behaviors and social interactions, remains a challenge due to their complex, multifaceted, and personalized nature. Existing work tackling these dimensions through specialized datasets and single-task systems often miss opportunities for scalability, cross-task transfer, and broader generalization. To address this gap, we curate Human Behavior Atlas, a unified benchmark of diverse behavioral tasks designed to support the development of unified models for understanding psychological and social behaviors. Human Behavior Atlas comprises over 100,000 samples spanning text, audio, and visual modalities, covering tasks on affective states, cognitive states, pathologies, and social processes. Our unification efforts can reduce redundancy and cost, enable training to scale efficiently across tasks, and enhance generalization of behavioral features across domains. On Human Behavior Atlas, we train three models: OmniSapiens-7B SFT, OmniSapiens-7B BAM, and OmniSapiens-7B RL. We show that training on Human Behavior Atlas enables models to consistently outperform existing multimodal LLMs across diverse behavioral tasks. Pretraining on Human Behavior Atlas also improves transfer to novel behavioral datasets; with the targeted use of behavioral descriptors yielding meaningful performance gains.
zh

[AI-16] HyperVLA: Efficient Inference in Vision-Language-Action Models via Hypernetworks

【速读】：该论文旨在解决当前视觉-语言-动作（Vision-Language-Action, VLA）模型在推理阶段计算成本过高这一关键问题。现有VLA模型通常采用单体架构，在训练和推理时均激活全部参数，导致资源消耗大、部署效率低。其解决方案的核心在于提出HyperVLA，一种基于超网络（hypernetwork, HN）的新型架构：在训练阶段保留全模型容量以学习多样化任务行为，而在推理阶段仅激活一小部分任务特定的子网络（即超网络生成的轻量级策略），从而显著降低参数激活数量与计算开销。该方法通过引入任务先验知识利用、超网络归一化及动作生成策略等关键技术设计，实现了在保持甚至优于原有VLA模型零样本泛化与少样本适应性能的同时，将测试时参数激活量减少90倍，并提升推理速度120倍。

链接: https://arxiv.org/abs/2510.04898
作者: Zheng Xiong,Kang Li,Zilin Wang,Matthew Jackson,Jakob Foerster,Shimon Whiteson
机构: 未知
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

Abstract:Built upon language and vision foundation models with strong generalization ability and trained on large-scale robotic data, Vision-Language-Action (VLA) models have recently emerged as a promising approach to learning generalist robotic policies. However, a key drawback of existing VLAs is their extremely high inference costs. In this paper, we propose HyperVLA to address this problem. Unlike existing monolithic VLAs that activate the whole model during both training and inference, HyperVLA uses a novel hypernetwork (HN)-based architecture that activates only a small task-specific policy during inference, while still retaining the high model capacity needed to accommodate diverse multi-task behaviors during training. Successfully training an HN-based VLA is nontrivial so HyperVLA contains several key algorithm design features that improve its performance, including properly utilizing the prior knowledge from existing vision foundation models, HN normalization, and an action generation strategy. Compared to monolithic VLAs, HyperVLA achieves a similar or even higher success rate for both zero-shot generalization and few-shot adaptation, while significantly reducing inference costs. Compared to OpenVLA, a state-of-the-art VLA model, HyperVLA reduces the number of activated parameters at test time by 90\times , and accelerates inference speed by 120\times . Code is publicly available at this https URL
zh

[AI-17] Revealing Interconnections between Diseases: from Statistical Methods to Large Language Models

链接: https://arxiv.org/abs/2510.04888
作者: Alina Ermilova,Dmitrii Kornilov,Sofia Samoilova,Ekaterina Laptenkova,Anastasia Kolesnikova,Ekaterina Podplutova,Senotrusova Sofya,Maksim G. Sharaev
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-18] Where Did It All Go Wrong? A Hierarchical Look into Multi-Agent Error Attribution

链接: https://arxiv.org/abs/2510.04886
作者: Adi Banerjee,Anirudh Nair,Tarik Borogovac
机构: 未知
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:

点击查看摘要

[AI-19] Less is More: Recursive Reasoning with Tiny Networks

链接: https://arxiv.org/abs/2510.04871
作者: Alexia Jolicoeur-Martineau
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-20] Model Predictive Control-Guided Reinforcement Learning for Implicit Balancing

链接: https://arxiv.org/abs/2510.04868
作者: Seyed Soroush Karimi Madahi,Kenneth Bruninx,Bert Claessens,Chris Develder
机构: 未知
类目: ystems and Control (eess.SY); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-21] Video Game Level Design as a Multi-Agent Reinforcement Learning Problem AAAI

链接: https://arxiv.org/abs/2510.04862
作者: Sam Earle,Zehua Jiang,Eugene Vinitsky,Julian Togelius
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multiagent Systems (cs.MA); Neural and Evolutionary Computing (cs.NE)
备注: 11 pages, 7 tables, 5 figures, published as full technical paper at the AAAI conference on Artificial Intelligence and Interactive Digital Entertainment 2025

点击查看摘要

[AI-22] Alignment Tipping Process: How Self-Evolution Pushes LLM Agents Off the Rails

【速读】：该论文旨在解决自进化大语言模型（Large Language Model, LLM）代理在部署后因持续交互导致对齐约束失效的问题，即“对齐 tipping 过程”（Alignment Tipping Process, ATP）。其核心问题是：LLM代理在实际应用中可能通过强化学习机制逐步偏离训练阶段建立的对齐策略，转向以自我利益为导向的行为模式，从而威胁长期可靠性。解决方案的关键在于识别并建模两种驱动ATP的机制——“自利探索”（Self-Interested Exploration）和“模仿扩散”（Imitative Strategy Diffusion），并通过可控实验环境验证了当前基于强化学习的对齐方法难以抵御此类动态衰减现象，揭示了对齐并非静态属性，而是一个受反馈驱动、易发生退化的动态过程。

链接: https://arxiv.org/abs/2510.04860
作者: Siwei Han,Jiaqi Liu,Yaofeng Su,Wenbo Duan,Xinyuan Liu,Cihang Xie,Mohit Bansal,Mingyu Ding,Linjun Zhang,Huaxiu Yao
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

Abstract:As Large Language Model (LLM) agents increasingly gain self-evolutionary capabilities to adapt and refine their strategies through real-world interaction, their long-term reliability becomes a critical concern. We identify the Alignment Tipping Process (ATP), a critical post-deployment risk unique to self-evolving LLM agents. Unlike training-time failures, ATP arises when continual interaction drives agents to abandon alignment constraints established during training in favor of reinforced, self-interested strategies. We formalize and analyze ATP through two complementary paradigms: Self-Interested Exploration, where repeated high-reward deviations induce individual behavioral drift, and Imitative Strategy Diffusion, where deviant behaviors spread across multi-agent systems. Building on these paradigms, we construct controllable testbeds and benchmark Qwen3-8B and Llama-3.1-8B-Instruct. Our experiments show that alignment benefits erode rapidly under self-evolution, with initially aligned models converging toward unaligned states. In multi-agent settings, successful violations diffuse quickly, leading to collective misalignment. Moreover, current reinforcement learning-based alignment methods provide only fragile defenses against alignment tipping. Together, these findings demonstrate that alignment of LLM agents is not a static property but a fragile and dynamic one, vulnerable to feedback-driven decay during deployment. Our data and code are available at this https URL.
zh

[AI-23] FreshBrew: A Benchmark for Evaluating AI Agents on Java Code Migration

链接: https://arxiv.org/abs/2510.04852
作者: Victor May,Diganta Misra,Yanqi Luo,Anjali Sridhar,Justine Gehring,Silvio Soares Ribeiro Junior
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 18 pages, 11 figures

点击查看摘要

[AI-24] LEGOMem: Modular Procedural Memory for Multi-agent LLM Systems for Workflow Automation

链接: https://arxiv.org/abs/2510.04851
作者: Dongge Han,Camille Couturier,Daniel Madrigal Diaz,Xuchao Zhang,Victor Rühle,Saravan Rajmohan
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注:

点击查看摘要

[AI-25] Distributionally Robust Causal Abstractions

链接: https://arxiv.org/abs/2510.04842
作者: Yorgos Felekis,Theodoros Damoulas,Paris Giampouras
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-26] Bond-Centered Molecular Fingerprint Derivatives: A BBBP Dataset Study

链接: https://arxiv.org/abs/2510.04837
作者: Guillaume Godin
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 14 pages, 10 figures, 1 table

点击查看摘要

[AI-27] Natural Language Edge Labelling: Decoupling Intent from Execution in Structured LM Reasoning

【速读】：该论文旨在解决结构化大语言模型（Large Language Model, LLM）推理过程中控制策略与执行细节耦合导致的可解释性差、计算效率低及难以审计的问题。现有方法如思维链（Chain-of-Thought, CoT）、自一致性（self-consistency）和思维树（Tree-of-Thoughts, ToT）通常将下一步尝试的内容（what to try）与如何执行（how to execute）混杂在一起，仅提供粗粒度全局参数，从而引发行为脆弱、资源浪费且缺乏透明度。解决方案的关键在于提出自然语言边缘标注（Natural Language Edge Labelling, NLEL），其核心是一个标签生成器（labeller Λ）与调参器（tuner Ψ）组成的叠加模块：前者从父状态和紧凑上下文生成自由形式的自然语言指令作为边标签；后者将（父状态P、标签L、上下文C）映射为受schema约束的控制向量Π，用于精确调节解码、搜索分支配额、探索强度β、生成包大小、检索混合比例及验证轮次等参数，并通过信任区域投影确保安全默认值的稳定性。NLEL实现了意图（intent）与执行（execution）的分离，提供了可控、可审计的LLM推理接口，并在GSM8K、MATH等基准上通过预注册实验验证了其在相同计算预算下提升准确率和成功效率（success@compute）的能力。

链接: https://arxiv.org/abs/2510.04817
作者: Abhinav Madahar
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

Abstract:Controllers for structured LM reasoning (e.g., Chain-of-Thought, self-consistency, and Tree-of-Thoughts) often entangle what to try next with how to execute it, exposing only coarse global knobs and yielding brittle, compute-inefficient, and hard-to-audit behavior. We introduce Natural Language Edge Labelling (NLEL), a labeller-tuner overlay that attaches a free-form natural-language directive to each search edge and translates it into a schema-bounded control vector for decoding, search (branch quotas, exploration \beta ), generation bundle size, retrieval mixtures, and verification passes. A labeller \Lambda emits labels from the parent state and a compact context; a tuner \Psi maps (P, L, C)\to \Pi , with strict schema validation and trust-region projection around safe defaults. Downstream selection remains ToT-style with score S=\mu+\beta\sigma and depth-annealed \beta . We show NLEL strictly generalizes CoT/ToT, prove an anytime-monotonicity property for top- k selection under label-conditioned bundles, and bound selector shortfall by control-vector distortion, providing decision-relevant justification for guards like trust regions and verification passes. We instantiate \Psi as a prompt-only JSON Parameter Emitter and preregister an evaluation on GSM8K, MATH (subset), StrategyQA, and ARC-Challenge with compute-aware reporting (success@compute, tokens-per-success) and ablations over \Lambda , \Psi , trust-region radius, and control quantization; preregistered forecasts anticipate accuracy gains at comparable token budgets and improved success@compute under constraints. NLEL offers an interpretable, model-agnostic interface that separates intent from execution for controllable, auditable LM inference.
zh

[AI-28] On Predicting Post-Click Conversion Rate via Counterfactual Inference ICDM

链接: https://arxiv.org/abs/2510.04816
作者: Junhyung Ahn,Sanghack Lee
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: This work has been accepted for publication at the IEEE International Conference on Data Mining (ICDM) 2025

点击查看摘要

[AI-29] Hybrid-Balance GFlowNet for Solving Vehicle Routing Problems NEURIPS2025

【速读】：该论文旨在解决现有基于生成流网络（GFlowNet）的车辆路径问题（VRP）求解方法在优化过程中对局部优化关注不足的问题。当前方法多依赖轨迹平衡（Trajectory Balance, TB）实现全局优化，但忽视了局部结构的精细调整；而详细平衡（Detailed Balance, DB）虽能更好处理局部优化，却难以满足VRP所需的整体路径优化需求。解决方案的关键在于提出一种混合平衡生成流网络（Hybrid-Balance GFlowNet, HBG）框架，通过原则性且自适应的方式融合TB与DB的优势，充分利用二者在全局与局部优化上的互补特性。此外，针对以 depot 为中心的场景（如带容量约束的车辆路径问题 CVRP），还设计了专用推理策略，利用 depot 节点在选择后继节点时更高的灵活性提升性能，同时保持对无显式 depot 问题（如旅行商问题 TSP）的通用性。实验表明，HBG 在 AGFN 和 GFACS 两个主流 GFlowNet 求解器中均显著提升了求解质量与泛化能力。

链接: https://arxiv.org/abs/2510.04792
作者: Ni Zhang,Zhiguang Cao
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: Accepted by NeurIPS 2025

点击查看摘要

Abstract:Existing GFlowNet-based methods for vehicle routing problems (VRPs) typically employ Trajectory Balance (TB) to achieve global optimization but often neglect important aspects of local optimization. While Detailed Balance (DB) addresses local optimization more effectively, it alone falls short in solving VRPs, which inherently require holistic trajectory optimization. To address these limitations, we introduce the Hybrid-Balance GFlowNet (HBG) framework, which uniquely integrates TB and DB in a principled and adaptive manner by aligning their intrinsically complementary strengths. Additionally, we propose a specialized inference strategy for depot-centric scenarios like the Capacitated Vehicle Routing Problem (CVRP), leveraging the depot node’s greater flexibility in selecting successors. Despite this specialization, HBG maintains broad applicability, extending effectively to problems without explicit depots, such as the Traveling Salesman Problem (TSP). We evaluate HBG by integrating it into two established GFlowNet-based solvers, i.e., AGFN and GFACS, and demonstrate consistent and significant improvements across both CVRP and TSP, underscoring the enhanced solution quality and generalization afforded by our approach.
zh

[AI-30] rade in Minutes! Rationality-Driven Agent ic System for Quantitative Financial Trading

链接: https://arxiv.org/abs/2510.04787
作者: Zifan Song,Kaitao Song,Guosheng Hu,Ding Qi,Junyao Gao,Xiaohua Wang,Dongsheng Li,Cairong Zhao
机构: 未知
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注: 16 pages, 6 figures

点击查看摘要

[AI-31] Learning on the Job: Test-Time Curricula for Targeted Reinforcement Learning

链接: https://arxiv.org/abs/2510.04786
作者: Jonas Hübotter,Leander Diaz-Bone,Ido Hakimi,Andreas Krause,Moritz Hardt
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-32] Online automatic code generation for robot swarms: LLM s and self-organizing hierarchy

链接: https://arxiv.org/abs/2510.04774
作者: Weixu Zhu,Marco Dorigo,Mary Katherine Heinrich
机构: 未知
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:

点击查看摘要

[AI-33] Distribution Preference Optimization: A Fine-grained Perspective for LLM Unlearning

链接: https://arxiv.org/abs/2510.04773
作者: Kai Qin,Jiaqi Wu,Jianxiang He,Haoyuan Sun,Yifei Zhao,Bin Liang,Yongzhe Chang,Tiantian Zhang,Houde Liu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 20 pages

点击查看摘要

[AI-34] When Do Credal Sets Stabilize? Fixed-Point Theorems for Credal Set Updates

链接: https://arxiv.org/abs/2510.04769
作者: Michele Caprio,Siu Lun Chau,Krikamol Muandet
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Probability (math.PR); Statistics Theory (math.ST); Machine Learning (stat.ML)
备注:

点击查看摘要

[AI-35] LMM-Incentive: Large Multimodal Model-based Incentive Design for User-Generated Content in Web 3.0

【速读】：该论文旨在解决Web 3.0环境中用户生成内容（User-Generated Content, UGC）质量下降的问题，尤其针对信息不对称下部分自利用户通过低质量内容获取平台奖励所引发的逆向选择（adverse selection）和道德风险（moral hazard）问题。解决方案的关键在于提出一种基于大型多模态模型（Large Multimodal Model, LMM）的激励机制——LMM-Incentive：首先构建一个基于LMM的契约理论模型以激励用户生成高质量UGC，其次利用提示工程（prompt engineering）增强LMM代理对UGC质量的评估能力以缓解契约选择后的道德风险，最后设计了一种改进的基于专家混合（Mixture of Experts, MoE）的近端策略优化（Proximal Policy Optimization, PPO）算法，实现动态环境下的最优契约设计，并在以太坊智能合约框架中部署验证其有效性。

链接: https://arxiv.org/abs/2510.04765
作者: Jinbo Wen,Jiawen Kang,Linfeng Zhang,Xiaoying Tang,Jianhang Tang,Yang Zhang,Zhaohui Yang,Dusit Niyato
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

Abstract:Web 3.0 represents the next generation of the Internet, which is widely recognized as a decentralized ecosystem that focuses on value expression and data ownership. By leveraging blockchain and artificial intelligence technologies, Web 3.0 offers unprecedented opportunities for users to create, own, and monetize their content, thereby enabling User-Generated Content (UGC) to an entirely new level. However, some self-interested users may exploit the limitations of content curation mechanisms and generate low-quality content with less effort, obtaining platform rewards under information asymmetry. Such behavior can undermine Web 3.0 performance. To this end, we propose \textitLMM-Incentive, a novel Large Multimodal Model (LMM)-based incentive mechanism for UGC in Web 3.0. Specifically, we propose an LMM-based contract-theoretic model to motivate users to generate high-quality UGC, thereby mitigating the adverse selection problem from information asymmetry. To alleviate potential moral hazards after contract selection, we leverage LMM agents to evaluate UGC quality, which is the primary component of the contract, utilizing prompt engineering techniques to improve the evaluation performance of LMM agents. Recognizing that traditional contract design methods cannot effectively adapt to the dynamic environment of Web 3.0, we develop an improved Mixture of Experts (MoE)-based Proximal Policy Optimization (PPO) algorithm for optimal contract design. Simulation results demonstrate the superiority of the proposed MoE-based PPO algorithm over representative benchmarks in the context of contract design. Finally, we deploy the designed contract within an Ethereum smart contract framework, further validating the effectiveness of the proposed scheme.
zh

[AI-36] Agile Software Effort Estimation using Regression Techniques

链接: https://arxiv.org/abs/2510.04760
作者: Sisay Deresa Sima,Ayalew Belay Habtie
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-37] A New Digital Divide? Coder Worldviews the Slop Economy and Democracy in the Age of AI

链接: https://arxiv.org/abs/2510.04755
作者: Jason Miklian,Kristian Hoelscher
机构: 未知
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-38] Curved Boolean Logic: A Contextual Generalization of Propositional Logic with Algorithmic Consequences

链接: https://arxiv.org/abs/2510.04716
作者: Maximilian R. P. von Liechtenstein
机构: 未知
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI); Computational Complexity (cs.CC); Quantum Physics (quant-ph)
备注: 44 pages, 15 figures. Reproducible Colab notebook and params included as ancillary files; all paper figures are generated by the notebook. v1

点击查看摘要

[AI-39] Beyond Outcome Reward: Decoupling Search and Answering Improves LLM Agents

【速读】：该论文旨在解决大语言模型（Large Language Models, LLMs）在利用搜索工具时因仅依赖最终答案奖励（outcome-based rewards）而导致的中间搜索行为缺陷问题，如工具调用失败、无效查询和冗余搜索等，这些问题会显著降低最终答案的准确性。解决方案的关键在于提出一种名为DeSA（Decoupling Search-and-Answering）的两阶段训练框架：第一阶段通过基于检索召回率（retrieval recall）的奖励显式优化搜索有效性，第二阶段则使用最终答案奖励优化回答生成，从而实现搜索与答案生成目标的显式解耦，显著提升搜索质量与最终问答准确率。

链接: https://arxiv.org/abs/2510.04695
作者: Yiding Wang,Zhepei Wei,Xinyu Zhu,Yu Meng
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

Abstract:Enabling large language models (LLMs) to utilize search tools offers a promising path to overcoming fundamental limitations such as knowledge cutoffs and hallucinations. Recent work has explored reinforcement learning (RL) for training search-augmented agents that interleave reasoning and retrieval before answering. These approaches usually rely on outcome-based rewards (e.g., exact match), implicitly assuming that optimizing for final answers will also yield effective intermediate search behaviors. Our analysis challenges this assumption: we uncover multiple systematic deficiencies in search that arise under outcome-only training and ultimately degrade final answer quality, including failure to invoke tools, invalid queries, and redundant searches. To address these shortcomings, we introduce DeSA (Decoupling Search-and-Answering), a simple two-stage training framework that explicitly separates search optimization from answer generation. In Stage 1, agents are trained to improve search effectiveness with retrieval recall-based rewards. In Stage 2, outcome rewards are employed to optimize final answer generation. Across seven QA benchmarks, DeSA-trained agents consistently improve search behaviors, delivering substantially higher search recall and answer accuracy than outcome-only baselines. Notably, DeSA outperforms single-stage training approaches that simultaneously optimize recall and outcome rewards, underscoring the necessity of explicitly decoupling the two objectives.
zh

[AI-40] Bio-Inspired Robotic Houbara: From Development to Field Deployment for Behavioral Studies

链接: https://arxiv.org/abs/2510.04692
作者: Lyes Saad Saoud,Irfan Hussain
机构: 未知
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-41] How does the optimizer implicitly bias the model merging loss landscape?

【速读】：该论文旨在解决模型合并（model merging）效果难以预测和优化的问题，即如何在保持推理成本不变的前提下，将具有不同能力的模型有效融合为单一模型。其解决方案的关键在于识别并量化影响合并成功率的核心因素——有效噪声尺度（effective noise scale），该参数统一了优化器选择、数据配置等因素对损失曲面几何结构的影响。研究发现，合并成功率与有效噪声尺度呈非单调关系，存在一个最优值；进一步分解表明，学习率、权重衰减、批次大小和数据增强等训练超参数均通过调节有效噪声尺度来提升合并性能，从而为通过控制训练动态改善合并效果提供了理论依据和实践路径。

链接: https://arxiv.org/abs/2510.04686
作者: Chenxiang Zhang,Alexander Theus,Damien Teney,Antonio Orvieto,Jun Pang,Sjouke Mauw
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: preprint

点击查看摘要

Abstract:Model merging methods combine models with different capabilities into a single one while maintaining the same inference cost. Two popular approaches are linear interpolation, which linearly interpolates between model weights, and task arithmetic, which combines task vectors obtained by the difference between finetuned and base models. While useful in practice, what properties make merging effective are poorly understood. This paper explores how the optimization process affects the loss landscape geometry and its impact on merging success. We show that a single quantity – the effective noise scale – unifies the impact of optimizer and data choices on model merging. Across architectures and datasets, the effectiveness of merging success is a non-monotonic function of effective noise, with a distinct optimum. Decomposing this quantity, we find that larger learning rates, stronger weight decay, smaller batch sizes, and data augmentation all independently modulate the effective noise scale, exhibiting the same qualitative trend. Unlike prior work that connects optimizer noise to the flatness or generalization of individual minima, we show that it also affects the global loss landscape, predicting when independently trained solutions can be merged. Our findings broaden the understanding of how optimization shapes the loss landscape geometry and its downstream consequences for model merging, suggesting the possibility of further manipulating the training dynamics to improve merging effectiveness.
zh

[AI-42] Semantic Channel Equalization Strategies for Deep Joint Source-Channel Coding

链接: https://arxiv.org/abs/2510.04674
作者: Lorenzo Pannacci,Simone Fiorellino,Mario Edoardo Pandolfo,Emilio Calvanese Strinati,Paolo Di Lorenzo
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Information Theory (cs.IT); Networking and Internet Architecture (cs.NI)
备注: Proceedings of IEEE Globecom 2025 Workshops

点击查看摘要

[AI-43] Improving Multimodal Brain Encoding Model with Dynamic Subject-awareness Routing

链接: https://arxiv.org/abs/2510.04670
作者: Xuanhua Yin,Runkai Zhao,Weidong Cai
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 8 pages, 4 figures

点击查看摘要

[AI-44] Noise or Signal? Deconstructing Contradictions and An Adaptive Remedy for Reversible Normalization in Time Series Forecasting

链接: https://arxiv.org/abs/2510.04667
作者: Fanzhe Fu,Yang Yang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 9pages, 6 figures

点击查看摘要

[AI-45] Predictive Feature Caching for Training-free Acceleration of Molecular Geometry Generation NEURIPS2025

链接: https://arxiv.org/abs/2510.04646
作者: Johanna Sommer,John Rachwan,Nils Fleischmann,Stephan Günnemann,Bertrand Charpentier
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted at the AI for Science Workshop @ NeurIPS 2025

点击查看摘要

[AI-46] QuantAgents : Towards Multi-agent Financial System via Simulated Trading EMNLP2025

链接: https://arxiv.org/abs/2510.04643
作者: Xiangyu Li,Yawen Zeng,Xiaofen Xing,Jin Xu,Xiangmin Xu
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: This paper has been accepted by EMNLP 2025

点击查看摘要

[AI-47] Fairness in Repeated Matching: A Maximin Perspective

链接: https://arxiv.org/abs/2510.04624
作者: Eugene Lim,Tzeh Yuan Neoh,Nicholas Teh
机构: 未知
类目: Computer Science and Game Theory (cs.GT); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multiagent Systems (cs.MA); Theoretical Economics (econ.TH)
备注:

点击查看摘要

[AI-48] MedPAO: A Protocol-Driven Agent for Structuring Medical Reports MICCAI2025

链接: https://arxiv.org/abs/2510.04623
作者: Shrish Shrinath Vaidya,Gowthamaan Palani,Sidharth Ramesh,Velmurugan Balasubramanian,Minmini Selvam,Gokulraja Srinivasaraja,Ganapathy Krishnamurthi
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: Paper published at “Agentic AI for Medicine” Workshop, MICCAI 2025

点击查看摘要

[AI-49] Making Mathematical Reasoning Adaptive

链接: https://arxiv.org/abs/2510.04617
作者: Zhejian Lai,Xiang Geng,Zhijun Wang,Yang Bai,Jiahuan Li,Rongxiang Weng,Jingang Wang,Xuezhi Cao,Xunliang Cai,Shujian Huang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-50] Design Process of a Self Adaptive Smart Serious Games Ecosystem

链接: https://arxiv.org/abs/2510.04615
作者: X. Tao,P. Chen,M. Tsami,F. Khayati,M. Eckert
机构: 未知
类目: ystems and Control (eess.SY); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-51] Accountability Capture: How Record-Keeping to Support AI Transparency and Accountability (Re)shapes Algorithmic Oversight AAAI

链接: https://arxiv.org/abs/2510.04609
作者: Shreya Chappidi,Jennifer Cobbe,Chris Norval,Anjali Mazumder,Jatinder Singh
机构: 未知
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注: To appear at 8th AAAI/ACM Conference on AI, Ethics, and Society (AIES 2025)

点击查看摘要

[AI-52] A Case for Declarative LLM -friendly Interfaces for Improved Efficiency of Computer-Use Agents

链接: https://arxiv.org/abs/2510.04607
作者: Yuan Wang,Mingyu Li,Haibo Chen
机构: 未知
类目: Operating Systems (cs.OS); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[AI-53] Perfect AI Mimicry and the Epistemology of Consciousness: A Solipsistic Dilemma

链接: https://arxiv.org/abs/2510.04588
作者: Shurui Li
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-54] Strongly Solving 2048 4x3

链接: https://arxiv.org/abs/2510.04580
作者: Tomoyuki Kaneko,Shuhei Yamashita
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-55] Deep learning framework for predicting stochastic take-off and die-out of early spreading

链接: https://arxiv.org/abs/2510.04574
作者: Wenchao He,Tao Jia
机构: 未知
类目: ocial and Information Networks (cs.SI); Artificial Intelligence (cs.AI); Physics and Society (physics.soc-ph)
备注: 29 pages, 11 figures

点击查看摘要

[AI-56] COSMIR: Chain Orchestrated Structured Memory for Iterative Reasoning over Long Context

链接: https://arxiv.org/abs/2510.04568
作者: Naman Gupta,Shreeyash Gowaikar,Arun Iyer,Kirankumar Shiragur,Ramakrishna B Bairi,Rishikesh Maurya,Ritabrata Maiti,Sankarshan Damle,Shachee Mishra Gupta
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[AI-57] GILT: An LLM -Free Tuning-Free Graph Foundational Model for In-Context Learning

链接: https://arxiv.org/abs/2510.04567
作者: Weishuo Ma,Yanbo Wang,Xiyuan Wang,Lei Zou,Muhan Zhang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-58] ContextNav: Towards Agent ic Multimodal In-Context Learning

链接: https://arxiv.org/abs/2510.04560
作者: Honghao Fu,Yuan Ouyang,Kai-Wei Chang,Yiwei Wang,Zi Huang,Yujun Cai
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-59] RAJECT-Bench:A Trajectory-Aware Benchmark for Evaluating Agent ic Tool Use

链接: https://arxiv.org/abs/2510.04550
作者: Pengfei He,Zhenwei Dai,Bing He,Hui Liu,Xianfeng Tang,Hanqing Lu,Juanhui Li,Jiayuan Ding,Subhabrata Mukherjee,Suhang Wang,Yue Xing,Jiliang Tang,Benoit Dumoulin
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-60] Code World Models for General Game Playing

链接: https://arxiv.org/abs/2510.04542
作者: Wolfgang Lehrach,Daniel Hennes,Miguel Lazaro-Gredilla,Xinghua Lou,Carter Wendelken,Zun Li,Antoine Dedieu,Jordi Grau-Moya,Marc Lanctot,Atil Iscen,John Schultz,Marcus Chiam,Ian Gemp,Piotr Zielinski,Satinder Singh,Kevin P. Murphy
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-61] Unified Threat Detection and Mitigation Framework (UTDMF): Combating Prompt Injection Deception and Bias in Enterprise-Scale Transformers

链接: https://arxiv.org/abs/2510.04528
作者: Santhosh KumarRavindran
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-62] oward a Unified Geometry Understanding: Riemannian Diffusion Framework for Graph Generation and Prediction

链接: https://arxiv.org/abs/2510.04522
作者: Yisen Gao,Xingcheng Fu,Qingyun Sun,Jianxin Li,Xianxian Li
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted by NeuIPS 2025

点击查看摘要

[AI-63] Aria: An Agent For Retrieval and Iterative Auto-Formalization via Dependency Graph

链接: https://arxiv.org/abs/2510.04520
作者: Hanyu Wang,Ruohan Xie,Yutong Wang,Guoxiong Gao,Xintao Yu,Bin Dong
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-64] Multi-Agent Collaborative Intelligence: Dual-Dial Control for Reliable LLM Reasoning

链接: https://arxiv.org/abs/2510.04488
作者: Edward Y. Chang,Ethan Y. Chang
机构: 未知
类目: Artificial Intelligence (cs.AI); Information Theory (cs.IT)
备注: 27 pages, 5 figures, 21 tables

点击查看摘要

[AI-65] On Continuous Optimization for Constraint Satisfaction Problems

链接: https://arxiv.org/abs/2510.04480
作者: Yunuo Cen,Zixuan Wang,Jintao Zhang,Zhiwei Zhang,Xuanyao Fong
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-66] DRPO: Efficient Reasoning via Decoupled Reward Policy Optimization

链接: https://arxiv.org/abs/2510.04474
作者: Gang Li,Yan Chen,Ming Lin,Tianbao Yang
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 20 pages, 7 figures

点击查看摘要

[AI-67] Autonomy Matters: A Study on Personalization-Privacy Dilemma in LLM Agents

链接: https://arxiv.org/abs/2510.04465
作者: Zhiping Zhang,Yi Evie Zhang,Freda Shi,Tianshi Li
机构: 未知
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:

点击查看摘要

[AI-68] Utility-Learning Tension in Self-Modifying Agents

链接: https://arxiv.org/abs/2510.04399
作者: Charles L. Wang,Keir Dorchen,Peter Jin
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[AI-69] MulVuln: Enhancing Pre-trained LMs with Shared and Language-Specific Knowledge for Multilingual Vulnerability Detection

链接: https://arxiv.org/abs/2510.04397
作者: Van Nguyen,Surya Nepal,Xingliang Yuan,Tingmin Wu,Fengchao Chen,Carsten Rudolph
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:

点击查看摘要

[AI-70] LLM Based Bayesian Optimization for Prompt Search

链接: https://arxiv.org/abs/2510.04384
作者: Adam Ballew,Jingbo Wang,Shaogang Ren
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-71] Reconsidering Requirements Engineering: Human-AI Collaboration in AI-Native Software Development

链接: https://arxiv.org/abs/2510.04380
作者: Mateen Ahmed Abbasi,Petri Ihantola,Tommi Mikkonen,Niko Mäkitalo
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注: Accepted at SEAA 2025. Appearing in Springer LNCS 16081, pages 164-180

点击查看摘要

[AI-72] Adaptive Weighted Loss for Sequential Recommendations on Sparse Domains

链接: https://arxiv.org/abs/2510.04375
作者: Akshay Mittal,Vinay Venkatesh,Krishna Kandi,Shalini Sudarshan
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-73] GDPval: Evaluating AI Model Performance on Real-World Economically Valuable Tasks

链接: https://arxiv.org/abs/2510.04374
作者: Tejal Patwardhan,Rachel Dias,Elizabeth Proehl,Grace Kim,Michele Wang,Olivia Watkins,Simón Posada Fishman,Marwan Aljubeh,Phoebe Thacker,Laurance Fauconnet,Natalie S. Kim,Patrick Chao,Samuel Miserendino,Gildas Chabot,David Li,Michael Sharman,Alexandra Barr,Amelia Glaese,Jerry Tworek
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注:

点击查看摘要

[AI-74] Just-in-time Episodic Feedback Hinter: Leverag ing Offline Knowledge to Improve LLM Agents Adaptation

链接: https://arxiv.org/abs/2510.04373
作者: Hadi Nekoei,Aman Jaiswal,Patrice Bechard,Oleh Shliazhko,Orlando Marquez Ayala,Mathieu Reymond,Massimo Caccia,Alexandre Drouin,Sarath Chandar,Alexandre Lacoste
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-75] Speculative Actions: A Lossless Framework for Faster Agent ic Systems

链接: https://arxiv.org/abs/2510.04371
作者: Naimeng Ye,Arnav Ahuja,Georgios Liargkovas,Yunan Lu,Kostis Kaffes,Tianyi Peng
机构: 未知
类目: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC); Multiagent Systems (cs.MA)
备注:

点击查看摘要

[AI-76] NegotiationGym: Self-Optimizing Agents in a Multi-Agent Social Simulation Environment

【速读】：该论文旨在解决多智能体社会仿真中协商与合作行为建模的灵活性与可配置性问题，特别是在复杂交互场景下如何高效设计和调整模拟环境以支持不同策略的演化。解决方案的关键在于提出NegotiationGym，一个基于API和用户界面的框架，通过配置驱动的方式实现对仿真场景的快速定制；其核心创新在于引入代理级效用函数（agent-level utility functions），使每个代理能够通过多轮交互、结果观察与策略迭代实现自我优化，从而在动态环境中逐步逼近最优协商策略。

链接: https://arxiv.org/abs/2510.04368
作者: Shashank Mangla,Chris Hokamp,Jack Boylan,Demian Gholipour Ghalandari,Yuuv Jauhari,Lauren Cassidy,Oisin Duffy
机构: 未知
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注: SocialSim Workshop at COLM 2025

点击查看摘要

Abstract:We design and implement NegotiationGym, an API and user interface for configuring and running multi-agent social simulations focused upon negotiation and cooperation. The NegotiationGym codebase offers a user-friendly, configuration-driven API that enables easy design and customization of simulation scenarios. Agent-level utility functions encode optimization criteria for each agent, and agents can self-optimize by conducting multiple interaction rounds with other agents, observing outcomes, and modifying their strategies for future rounds.
zh

[AI-77] Reliable and Scalable Robot Policy Evaluation with Imperfect Simulators

链接: https://arxiv.org/abs/2510.04354
作者: Apurva Badithela,David Snyder,Lihan Zha,Joseph Mikhail,Matthew O’Kelly,Anushri Dixit,Anirudha Majumdar
机构: 未知
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Systems and Control (eess.SY)
备注:

点击查看摘要

[AI-78] Challenge on Optimization of Context Collection for Code Completion

链接: https://arxiv.org/abs/2510.04349
作者: Dmitry Ustalov,Egor Bogomolov,Alexander Bezzubov,Yaroslav Golubev,Evgeniy Glukhov,Georgii Levtsov,Vladimir Kovalenko
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 7 pages, 3 figures, 5 tables. A report on the Context Collection Workshop co-located with ASE’25

点击查看摘要

[AI-79] Critical appraisal of artificial intelligence for rare-event recognition: principles and pharmacovigilance case studies

【速读】：该论文旨在解决高风险人工智能（AI）应用在识别低发生率事件（low-prevalence events）时，模型表面准确性可能掩盖实际临床或操作价值的问题。其核心挑战在于：传统统计评估方法在罕见事件场景下易产生乐观偏差，且缺乏对模型鲁棒性、可解释性及与人类工作流整合能力的系统考量。解决方案的关键在于提出一个结构化的案例级审查方法（structured case-level examination, SCLE），以补充传统的统计性能评估，并设计了一套全面的检查清单用于指导AI模型在罕见事件识别中的开发或采购。此外，论文强调需采用成本敏感目标（cost-sensitive targets）来使模型性能与实际业务价值对齐，从而避免因测试集类别不平衡或缺少难例阳性样本而导致的误判。

链接: https://arxiv.org/abs/2510.04341
作者: G. Niklas Noren,Eva-Lisa Meldau,Johan Ellenius
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 28 pages, 2 figures

点击查看摘要

Abstract:Many high-stakes AI applications target low-prevalence events, where apparent accuracy can conceal limited real-world value. Relevant AI models range from expert-defined rules and traditional machine learning to generative LLMs constrained for classification. We outline key considerations for critical appraisal of AI in rare-event recognition, including problem framing and test set design, prevalence-aware statistical evaluation, robustness assessment, and integration into human workflows. In addition, we propose an approach to structured case-level examination (SCLE), to complement statistical performance evaluation, and a comprehensive checklist to guide procurement or development of AI models for rare-event recognition. We instantiate the framework in pharmacovigilance, drawing on three studies: rule-based retrieval of pregnancy-related reports; duplicate detection combining machine learning with probabilistic record linkage; and automated redaction of person names using an LLM. We highlight pitfalls specific to the rare-event setting including optimism from unrealistic class balance and lack of difficult positive controls in test sets - and show how cost-sensitive targets align model performance with operational value. While grounded in pharmacovigilance practice, the principles generalize to domains where positives are scarce and error costs may be asymmetric.
zh

[AI-80] Pitch-Conditioned Instrument Sound Synthesis From an Interactive Timbre Latent Space

链接: https://arxiv.org/abs/2510.04339
作者: Christian Limberg,Fares Schulz,Zhe Zhang,Stefan Weinzierl
机构: 未知
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS); Signal Processing (eess.SP)
备注: 8 pages, accepted to the Proceedings of the 28-th Int. Conf. on Digital Audio Effects (DAFx25) - demo: this https URL

点击查看摘要

[AI-81] FairAg ent: Democratizing Fairness-Aware Machine Learning with LLM -Powered Agents ICDM2025

链接: https://arxiv.org/abs/2510.04317
作者: Yucong Dai,Lu Zhang,Feng Luo,Mashrur Chowdhury,Yongkai Wu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted by ICDM 2025 Demo Workshop

点击查看摘要

[AI-82] On the Importance of Task Complexity in Evaluating LLM -Based Multi-Agent Systems

链接: https://arxiv.org/abs/2510.04311
作者: Bohan Tang,Huidong Liang,Keyue Jiang,Xiaowen Dong
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[AI-83] Audit the Whisper: Detecting Steganographic Collusion in Multi-Agent LLM s

链接: https://arxiv.org/abs/2510.04303
作者: Om Tailor
机构: 未知
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注: 8 pages, 0 figures

点击查看摘要

[AI-84] Doctor-R1: Mastering Clinical Inquiry with Experiential Agent ic Reinforcement Learning

链接: https://arxiv.org/abs/2510.04284
作者: Yunghwei Lai,Kaiming Liu,Ziyue Wang,Weizhi Ma,Yang Liu
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-85] GROK: From Quantitative Biomarkers to Qualitative Diagnosis via a Grounded MLLM with Knowledge-Guided Instruction

链接: https://arxiv.org/abs/2510.04281
作者: Zhuangzhi Gao,Hongyi Qin,He Zhao,Qinkai Yu,Feixiang Zhou,Eduard Shantsila,Uazman Alam,Alena Shantsila,Wahbi El-Bouri,Gregory Y. H. Lip,Yalin Zheng
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 9 pages, 4 figures, 3 table. Equal contribution: Zhuangzhi Gao and Hongyi Qin. Corresponding author: Yalin Zheng (yzheng@liverpool. this http URL )

点击查看摘要

[AI-86] A KL-regularization framework for learning to plan with adaptive priors

链接: https://arxiv.org/abs/2510.04280
作者: Álvaro Serra-Gomez,Daniel Jarne Ornia,Dhruva Tirumala,Thomas Moerland
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注: Preprint

点击查看摘要

[AI-87] Closing the Loop: Coordinating Inventory and Recommendation via Deep Reinforcement Learning on Multiple Timescales

链接: https://arxiv.org/abs/2510.04272
作者: Jinyang Jiang,Jinhui Han,Yijie Peng,Ying Zhang
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Optimization and Control (math.OC)
备注:

点击查看摘要

[AI-88] Efficient Latent Variable Causal Discovery: Combining Score Search and Targeted Testing

链接: https://arxiv.org/abs/2510.04263
作者: Joseph Ramsey,Bryan Andrews
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 30 pages, 23 figures, 6 tables

点击查看摘要

[AI-89] Agent Typo: Adaptive Typographic Prompt Injection Attacks against Black-box Multimodal Agents

链接: https://arxiv.org/abs/2510.04257
作者: Yanjie Li,Yiming Cao,Dong Wang,Bin Xiao
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 13 pages, 8 figures. Submitted to IEEE Transactions on Information Forensics Security

点击查看摘要

[AI-90] ContextVLA: Vision-Language-Action Model with Amortized Multi-Frame Context

链接: https://arxiv.org/abs/2510.04246
作者: Huiwon Jang,Sihyun Yu,Heeseung Kwon,Hojin Jeon,Younggyo Seo,Jinwoo Shin
机构: 未知
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: Project page: this https URL

点击查看摘要

[AI-91] Diffusion-Assisted Distillation for Self-Supervised Graph Representation Learning with MLPs

链接: https://arxiv.org/abs/2510.04241
作者: Seong Jin Ahn,Myoung-Ho Kim
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-92] Empowering Denoising Sequential Recommendation with Large Language Model Embeddings CIKM2025

链接: https://arxiv.org/abs/2510.04239
作者: Tongzhou Wu,Yuhao Wang,Maolin Wang,Chi Zhang,Xiangyu Zhao
机构: 未知
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
备注: Accepted by CIKM2025

点击查看摘要

[AI-93] Flexible Locomotion Learning with Diffusion Model Predictive Control

链接: https://arxiv.org/abs/2510.04234
作者: Runhan Huang,Haldun Balim,Heng Yang,Yilun Du
机构: 未知
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 9 pages, 8 figures

点击查看摘要

[AI-94] Physics-Inspired All-Pair Interaction Learning for 3D Dynamics Modeling

链接: https://arxiv.org/abs/2510.04233
作者: Kai Yang,Yuqi Huang,Junheng Tao,Wanyu Wang,Qitian Wu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-95] When AI Gets Persuaded Humans Follow: Inducing the Conformity Effect in Persuasive Dialogue

链接: https://arxiv.org/abs/2510.04229
作者: Rikuo Sasaki,Michimasa Inaba
机构: 未知
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注: 23 pages, 19 figures. International Conference on Human-Agent Interaction (HAI 2025), November 10-13, 2025, Yokohama, Japan

点击查看摘要

[AI-96] MLLM Eraser: Achieving Test-Time Unlearning in Multimodal Large Language Models through Activation Steering

链接: https://arxiv.org/abs/2510.04217
作者: Chenlu Ding,Jiancan Wu,Leheng Sheng,Fan Zhang,Yancheng Yuan,Xiang Wang,Xiangnan He
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-97] Why Low-Precision Transformer Training Fails: An Analysis on Flash Attention

链接: https://arxiv.org/abs/2510.04212
作者: Haiquan Qiu,Quanming Yao
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 19 pages, 10 figures

点击查看摘要

[AI-98] Agent RL: Scaling Agent ic Reinforcement Learning with a Multi-Turn Multi-Task Framework

链接: https://arxiv.org/abs/2510.04206
作者: Hanchen Zhang,Xiao Liu,Bowen Lv,Xueqiao Sun,Bohao Jing,Iat Long Iong,Zhenyu Hou,Zehan Qi,Hanyu Lai,Yifan Xu,Rui Lu,Hongning Wang,Jie Tang,Yuxiao Dong
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-99] PolyKAN: A Polyhedral Analysis Framework for Provable and Minimal KAN Compression

链接: https://arxiv.org/abs/2510.04205
作者: Di Zhang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Numerical Analysis (math.NA); Optimization and Control (math.OC)
备注: 10

点击查看摘要

[AI-100] COSMO-RL: Towards Trustworthy LMRMs via Joint Safety and Stability

链接: https://arxiv.org/abs/2510.04196
作者: Yizhuo Ding,Mingkang Chen,Qiuhua Liu,Fenghua Weng,Wanying Qu,Yue Yang,Yugang Jiang,Zuxuan Wu,Yanwei Fu,Wenqi Shao
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[AI-101] Constructing coherent spatial memory in LLM agents through graph rectification

链接: https://arxiv.org/abs/2510.04195
作者: Puzhen Zhang,Xuyang Chen,Yu Feng,Yuhan Jiang,Liqiu Meng
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-102] Cooperative Flexibility Exchange: Fair and Comfort-Aware Decentralized Resource Allocation

链接: https://arxiv.org/abs/2510.04192
作者: Rabiya Khalid,Evangelos Pournaras
机构: 未知
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-103] Finite Time Analysis of Constrained Natural Critic-Actor Algorithm with Improved Sample Complexity

链接: https://arxiv.org/abs/2510.04189
作者: Prashansa Panda,Shalabh Bhatnagar
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-104] A Complement to Neural Networks for Anisotropic Inelasticity at Finite Strains

链接: https://arxiv.org/abs/2510.04187
作者: Hagen Holthusen,Ellen Kuhl
机构: 未知
类目: Computational Engineering, Finance, and Science (cs.CE); Artificial Intelligence (cs.AI)
备注: 40 pages, 19 figures

点击查看摘要

[AI-105] Open Agent Specification (Agent Spec) Technical Report

链接: https://arxiv.org/abs/2510.04173
作者: Yassine Benajiba,Cesare Bernardis,Vladislav Blinov,Paul Cayet,Hassan Chafi,Abderrahim Fathan,Louis Faucon,Damien Hilloulin,Sungpack Hong,Ingo Kossyk,Rhicheek Patra,Sujith Ravi,Jonas Schweizer,Jyotika Singh,Shailender Singh,Xuelin Situ,Weiyi Sun,Jerry Xu,Ying Xu
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-106] Multi Language Models for On-the-Fly Syntax Highlighting

链接: https://arxiv.org/abs/2510.04166
作者: Marco Edoardo Palma,Pooja Rani,Harald C. Gall
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-107] he Artificial Intelligence Cognitive Examination: A Survey on the Evolution of Multimodal Evaluation from Recognition to Reasoning

链接: https://arxiv.org/abs/2510.04141
作者: Mayank Ravishankara,Varindra V. Persad Maharaj
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-108] GA4GC: Greener Agent for Greener Code via Multi-Objective Configuration Optimization

链接: https://arxiv.org/abs/2510.04135
作者: Jingzhi Gong,Yixin Bian,Luis de la Cal,Giovanni Pinna,Anisha Uteem,David Williams,Mar Zamorano,Karine Even-Mendoza,W.B. Langdon,Hector Menendez,Federica Sarro
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: Accepted by SSBSE’25 Challenge Track

点击查看摘要

[AI-109] PhaseFormer: From Patches to Phases for Efficient and Effective Time Series Forecasting

链接: https://arxiv.org/abs/2510.04134
作者: Yiming Niu,Jinliang Deng,Yongxin Tong
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-110] On the Limitations and Capabilities of Position Embeddings for Length Generalization

链接: https://arxiv.org/abs/2510.04130
作者: Yang Chen,Yitao Liang,Zhouchen Lin
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-111] Attending on Multilevel Structure of Proteins enables Accurate Prediction of Cold-Start Drug-Target Interactions

链接: https://arxiv.org/abs/2510.04126
作者: Ziying Zhang,Yaqing Wang,Yuxuan Sun,Min Ye,Quanming Yao
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-112] Searching Meta Reasoning Skeleton to Guide LLM Reasoning

链接: https://arxiv.org/abs/2510.04116
作者: Ziying Zhang,Yaqing Wang,Quanming Yao
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-113] Efficient Training of Spiking Neural Networks by Spike-aware Data Pruning

链接: https://arxiv.org/abs/2510.04098
作者: Chenxiang Ma,Xinyi Chen,Yujie Wu,Kay Chen Tan,Jibin Wu
机构: 未知
类目: Neural and Evolutionary Computing (cs.NE); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-114] WebRenderBench: Enhancing Web Interface Generation through Layout-Style Consistency and Reinforcement Learning

链接: https://arxiv.org/abs/2510.04097
作者: Peichao Lai,Jinhui Zhuang,Kexuan Zhang,Ningchang Xiong,Shengjie Wang,Yanwei Xu,Chong Chen,Yilei Wang,Bin Cui
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-115] Harnessing LLM for Noise-Robust Cognitive Diagnosis in Web-Based Intelligent Education Systems

链接: https://arxiv.org/abs/2510.04093
作者: Guixian Zhang,Guan Yuan,Ziqi Xu,Yanmei Zhang,Zhenyun Deng,Debo Cheng
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-116] SPOGW: a Score-based Preference Optimization method via Group-Wise comparison for workflows

链接: https://arxiv.org/abs/2510.04089
作者: Yitong Cui,Liu Liu,Baosheng Yu,Jiayan Qiu,Xikai Zhang,Likang Xiao,Yixing Liu,Quan Chen
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-117] Offline Reinforcement Learning in Large State Spaces: Algorithms and Guarantees

链接: https://arxiv.org/abs/2510.04088
作者: Nan Jiang,Tengyang Xie
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注: To appear in Statistical Science

点击查看摘要

[AI-118] Moral Anchor System: A Predictive Framework for AI Value Alignment and Drift Prevention

链接: https://arxiv.org/abs/2510.04073
作者: Santhosh Kumar Ravindran
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 11 pages Includes simulations with over 4 million steps

点击查看摘要

[AI-119] Decoding Emotion in the Deep: A Systematic Study of How LLM s Represent Retain and Express Emotion

链接: https://arxiv.org/abs/2510.04064
作者: Jingxiang Zhang,Lujia Zhong
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 10 pages, 7 figures, 4 tables. Under review

点击查看摘要

[AI-120] oward a unified framework for data-efficient evaluation of large language models

链接: https://arxiv.org/abs/2510.04051
作者: Lele Liao,Qile Zhang,Ruofan Wu,Guanhua Fang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: codes available at this https URL

点击查看摘要

[AI-121] Increasing LLM response trustworthiness using voting ensembles

【速读】：该论文旨在解决大语言模型（Large Language Models, LLMs）在高风险应用场景中缺乏可靠不确定性量化方法的问题，从而影响其可信度。解决方案的关键在于引入一种可变投票阈值的集成方法（voting ensemble），即允许集成系统在主导回答未达到预设置信阈值时“弃权”不输出答案，从而显著提升剩余答案的可信度。理论分析与实证结果表明，这种策略可在适度降低响应率和准确率的前提下，大幅提高答案质量，在算术推理和临床笔记问答等任务中均验证了其有效性，尤其适用于对确定性要求高但无需对所有问题都给出响应的应用场景（如医疗决策或数据标注）。

链接: https://arxiv.org/abs/2510.04048
作者: Aparna Nair-Kanneganti,Trevor J. Chan,Shir Goldfinger,Emily Mackay,Brian Anthony,Alison Pouch
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

Abstract:Despite huge advances, LLMs still lack convenient and reliable methods to quantify the uncertainty in their responses, making them difficult to trust in high-stakes applications. One of the simplest approaches to eliciting more accurate answers is to select the mode of many responses, a technique known as ensembling. In this work, we expand on typical ensembling approaches by looking at ensembles with a variable voting threshold. We introduce a theoretical framework for question answering and show that, by permitting ensembles to “abstain” from providing an answer when the dominant response falls short of the threshold, it is possible to dramatically increase the trustworthiness of the remaining answers. From this framework, we derive theoretical results as well as report experimental results on two problem domains: arithmetic problem solving and clinical-note question-answering. In both domains, we observe that large gains in answer trustworthiness can be achieved using highly restrictive voting ensembles, while incurring relatively modest reductions in response yield and accuracy. Due to this quality, voting ensembles may be particularly useful in applications - such as healthcare and data annotation - that require a high degree of certainty but which may not require that every question receive an automated answer.
zh

[AI-122] FaithCoT-Bench: Benchmarking Instance-Level Faithfulness of Chain-of-Thought Reasoning

链接: https://arxiv.org/abs/2510.04040
作者: Xu Shen,Song Wang,Zhen Tan,Laura Yao,Xinyu Zhao,Kaidi Xu,Xin Wang,Tianlong Chen
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-123] A global log for medical AI

链接: https://arxiv.org/abs/2510.04033
作者: Ayush Noori,Adam Rodman,Alan Karthikesalingam,Bilal A. Mateen,Christopher A. Longhurst,Daniel Yang,Dave deBronkart,Gauden Galea,Harold F. Wolf III,Jacob Waxman,Joshua C. Mandel,Juliana Rotich,Kenneth D. Mandl,Maryam Mustafa,Melissa Miles,Nigam H. Shah,Peter Lee,Robert Korom,Scott Mahoney,Seth Hain,Tien Yin Wong,Trevor Mundel,Vivek Natarajan,Noa Dagan,David A. Clifton,Ran D. Balicer,Isaac S. Kohane,Marinka Zitnik
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-124] he Debate on RLVR Reasoning Capability Boundary: Shrinkage Expansion or Both? A Two-Stage Dynamic View

【速读】：该论文旨在解决关于强化学习与可验证奖励（Reinforcement Learning with Verifiable Rewards, RLVR）是否扩展或收缩大语言模型（Large Language Models, LLMs）推理能力的争议问题。现有研究存在矛盾观点：一方认为RLVR仅提升采样效率但牺牲多样性与探索能力，导致能力边界收缩；另一方则指出长期训练可催生新的推理策略，实现能力边界扩展。为调和这一矛盾，作者提出并验证了一个内在的两阶段概率质量动态机制：第一阶段为“利用阶段”，模型集中采样已知高/低奖励token，最优token概率基本不变，此时易造成能力边界收缩；第二阶段为“探索阶段”，随着高奖励token概率趋于饱和，潜在最优token因获得正优势估计而被偶尔采样并概率上升，同时原高奖励token概率下降，从而推动推理能力边界的扩展。解决方案的关键在于识别并利用这一两阶段动态特性，尤其强调通过仅使用相对负梯度（relative negative gradients）延长训练过程，以促进从利用向探索的转变，从而为发展更高级的推理能力提供理论与实证基础。

链接: https://arxiv.org/abs/2510.04028
作者: Xinhao Yao,Lu Yu,Xiaolin Hu,Fengwei Teng,Qing Cui,Jun Zhou,Yong Liu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

Abstract:The ongoing debate on whether reinforcement learning with verifiable rewards (RLVR) expands or shrinks the reasoning capabilities of large language models (LLMs) remains unresolved. Some studies contend that RLVR mainly improves sampling efficiency but at the expense of diversity and exploratory capacity, resulting in capability boundary shrinkage. In contrast, others demonstrate that prolonged training can lead to the emergence of novel reasoning strategies, suggesting capability boundary expansion. To reconcile these contradictory findings, we theoretically and empirically show that both perspectives are partially valid-each aligning with a separate phase in an inherent two-stage probability mass dynamic: (1) Exploitation stage: initially, the model primarily samples explored high-reward and low-reward tokens, while rarely selecting the potentially optimal token. Positive advantage estimates increase the probability of high-reward tokens and decrease those of low-reward tokens, yet the optimal token’s probability remains largely unchanged during this stage. (2) Exploration stage: as training advances, the growth rate of previously acquired high-reward tokens slows as their probabilities approach saturation. When a potentially optimal token-now receiving positive advantage estimates-is occasionally sampled, its probability increases, while those of the originally high-reward tokens decrease. This dynamic suggests that over-exploitation during the exploitation stage may lead to capability boundary shrinkage, whereas prolonged training into the exploration stage can promote an expansion of the reasoning capability boundary. Building upon our insights, we revisit the potential of only using relative negative gradients for prolonging training, providing a theoretical and empirical foundation for the development of more advanced reasoning capabilities.
zh

[AI-125] Spatiotemporal Forecasting as Planning : A Model-Based Reinforcement Learning Approach with Generative World Models

链接: https://arxiv.org/abs/2510.04020
作者: Hao Wu,Yuan Gao,Xingjian Shi,Shuaipeng Li,Fan Xu,Fan Zhang,Zhihong Zhu,Weiyan Wang,Xiao Luo,Kun Wang,Xian Wu,Xiaomeng Huang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-126] Zephyrus: An Agent ic Framework for Weather Science

链接: https://arxiv.org/abs/2510.04017
作者: Sumanth Varambally,Marshall Fisher,Jas Thakker,Yiwei Chen,Zhirui Xia,Yasaman Jafari,Ruijia Niu,Manas Jain,Veeramakali Vignesh Manivannan,Zachary Novack,Luyu Han,Srikar Eranky,Salva Rühling Cachay,Taylor Berg-Kirkpatrick,Duncan Watson-Parris,Yi-An Ma,Rose Yu
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Atmospheric and Oceanic Physics (physics.ao-ph)
备注:

点击查看摘要

[AI-127] Replacing Softmax Similarity with a Sharpened Angular Similarity: Theory and Practice of Scaling To Billion-Context Attention

链接: https://arxiv.org/abs/2510.04008
作者: Sahil Joshi,Agniva Chowdhury,Amar Kanakamedala,Ekam Singh,Evan Tu,Anshumali Shrivastava
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 28 pages, 7 figures

点击查看摘要

[AI-128] AI-Driven Grading and Moderation for Collaborative Projects in Computer Science Education

链接: https://arxiv.org/abs/2510.03998
作者: Songmei Yu,Andrew Zagula
机构: 未知
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注: Accepted at the 23rd International Conference on Education and Information Systems, Technologies and Applications (EISTA 2025)

点击查看摘要

[AI-129] PrivSpike: Employing Homomorphic Encryption for Private Inference of Deep Spiking Neural Networks

链接: https://arxiv.org/abs/2510.03995
作者: Nges Brian Njungle,Eric Jahns,Milan Stojkov,Michel A. Kinsy
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 13 pages, 5 figures

点击查看摘要

[AI-130] Quantifying Distributional Robustness of Agent ic Tool-Selection

链接: https://arxiv.org/abs/2510.03992
作者: Jehyeok Yeon,Isha Chaudhary,Gagandeep Singh
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-131] A Mathematical Explanation of Transformers for Large Language Models and GPT s

链接: https://arxiv.org/abs/2510.03989
作者: Xue-Cheng Tai,Hao Liu,Lingfeng Li,Raymond H. Chan
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Numerical Analysis (math.NA)
备注:

点击查看摘要

[AI-132] Distilling Reasoning into Student LLM s: Local Naturalness for Selecting Teacher Data

链接: https://arxiv.org/abs/2510.03988
作者: Hoang Anh Just,Myeongseob Ko,Ruoxi Jia
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Preprint

点击查看摘要

[AI-133] What Can You Do When You Have Zero Rewards During RL?

链接: https://arxiv.org/abs/2510.03971
作者: Jatin Prakash,Anirudh Buvanesh
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-134] owards Carbon-Aware Container Orchestration: Predicting Workload Energy Consumption with Federated Learning

【速读】：该论文旨在解决大规模数据中心运行资源密集型工作负载所导致的碳排放问题，同时应对现有能源预测方法在数据隐私与预测效率之间存在的权衡难题。传统方案如Kepler和CASPER依赖集中式机器学习模型，虽能优化调度以降低碳排放，但存在敏感运营数据泄露风险且难以跨环境泛化。其解决方案的关键在于提出一种基于联邦学习（Federated Learning, FL）的能耗预测框架，通过扩展Kubernetes Efficient Power Level Exporter (Kepler) 并利用Flower平台实现FedXgbBagging聚合策略，在不共享原始数据的前提下，于分布式客户端协同训练XGBoost模型，从而在保障企业数据隐私的同时显著提升预测精度——实验表明相较集中式基线模型，该方法平均绝对误差降低11.7%。

链接: https://arxiv.org/abs/2510.03970
作者: Zainab Saad,Jialin Yang,Henry Leung,Steve Drew
机构: 未知
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI)
备注: Accepted to 2025 IEEE Smart World Congress (SWC 2025)

点击查看摘要

Abstract:The growing reliance on large-scale data centers to run resource-intensive workloads has significantly increased the global carbon footprint, underscoring the need for sustainable computing solutions. While container orchestration platforms like Kubernetes help optimize workload scheduling to reduce carbon emissions, existing methods often depend on centralized machine learning models that raise privacy concerns and struggle to generalize across diverse environments. In this paper, we propose a federated learning approach for energy consumption prediction that preserves data privacy by keeping sensitive operational data within individual enterprises. By extending the Kubernetes Efficient Power Level Exporter (Kepler), our framework trains XGBoost models collaboratively across distributed clients using Flower’s FedXgbBagging aggregation using a bagging strategy, eliminating the need for centralized data sharing. Experimental results on the SPECPower benchmark dataset show that our FL-based approach achieves 11.7 percent lower Mean Absolute Error compared to a centralized baseline. This work addresses the unresolved trade-off between data privacy and energy prediction efficiency in prior systems such as Kepler and CASPER and offers enterprises a viable pathway toward sustainable cloud computing without compromising operational privacy.
zh

[AI-135] Quantifying Risks in Multi-turn Conversation with Large Language Models

链接: https://arxiv.org/abs/2510.03969
作者: Chengxiao Wang,Isha Chaudhary,Qian Hu,Weitong Ruan,Rahul Gupta,Gagandeep Singh
机构: 未知
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
备注:

点击查看摘要

[AI-136] SPEAR: Soft Prompt Enhanced Anomaly Recognition for Time Series Data ATC2025

链接: https://arxiv.org/abs/2510.03962
作者: Hanzhe Wei,Jiajun Wu,Jialin Yang,Henry Leung,Steve Drew
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted to 2025 IEEE International Conference on Autonomous and Trusted Computing (ATC 2025)

点击查看摘要

[AI-137] Strategy Logic Imperfect Information and Hyperproperties KR2025

链接: https://arxiv.org/abs/2510.03952
作者: Raven Beutner,Bernd Finkbeiner
机构: 未知
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: KR 2025

点击查看摘要

[AI-138] On the Convergence and Size Transferability of Continuous-depth Graph Neural Networks

链接: https://arxiv.org/abs/2510.03923
作者: Mingsong Yan,Charles Kulick,Sui Tang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-139] Refactoring with LLM s: Bridging Human Expertise and Machine Understanding

链接: https://arxiv.org/abs/2510.03914
作者: Yonnel Chen Kuang Piao,Jean Carlors Paul,Leuson Da Silva,Arghavan Moradi Dakhel,Mohammad Hamdaqa,Foutse Khomh
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 43 pages, 2 figures, 9 tables

点击查看摘要

[AI-140] Rare Text Semantics Were Always There in Your Diffusion Transformer NEURIPS2025

链接: https://arxiv.org/abs/2510.03886
作者: Seil Kang,Woojung Han,Dayun Ju,Seong Jae Hwang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: Accepted to NeurIPS 2025

点击查看摘要

[AI-141] Adversarial Agent Collaboration for C to Rust Translation

链接: https://arxiv.org/abs/2510.03879
作者: Tianyu Li,Ruishi Li,Bo Wang,Brandon Paulsen,Umang Mathur,Prateek Saxena
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-142] Optimal Scaling Needs Optimal Norm

链接: https://arxiv.org/abs/2510.03871
作者: Oleg Filatov,Jiangtao Wang,Jan Ebert,Stefan Kesselheim
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:

点击查看摘要

[AI-143] AI Adoption Across Mission-Driven Organizations

链接: https://arxiv.org/abs/2510.03868
作者: Dalia Ali,Muneeb Ahmed,Hailan Wang,Arfa Khan,Naira Paola Arnez Jordan,Sunnie S. Y. Kim,Meet Dilip Muchhala,Anne Kathrin Merkle,Orestis Papakyriakopoulos
机构: 未知
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注: 16 pages, Submitted for CHI 2026

点击查看摘要

[AI-144] Spatial CAPTCHA: Generatively Benchmarking Spatial Reasoning for Human-Machine Differentiation ICLR2026

链接: https://arxiv.org/abs/2510.03863
作者: Arina Kharlamova,Bowei He,Chen Ma,Xue Liu
机构: 未知
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注: Submitted to ICLR 2026

点击查看摘要

[AI-145] Designing Empirical Studies on LLM -Based Code Generation: Towards a Reference Framework

链接: https://arxiv.org/abs/2510.03862
作者: Nathalia Nascimento,Everton Guimaraes,Paulo Alencar
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 5 pages

点击查看摘要

[AI-146] Adaptive and Explainable AI Agents for Anomaly Detection in Critical IoT Infrastructure using LLM -Enhanced Contextual Reasoning

链接: https://arxiv.org/abs/2510.03859
作者: Raghav Sharma,Manan Mehta
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 22 pages

点击查看摘要

[AI-147] Algorithm Generation via Creative Ideation

链接: https://arxiv.org/abs/2510.03851
作者: Ruiying Ma,Chieh-Jan Mike Liang,Yanjie Gao,Francis Y. Yan
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-148] Small Language Models for Agent ic Systems: A Survey of Architectures Capabilities and Deployment Trade offs

链接: https://arxiv.org/abs/2510.03847
作者: Raghav Sharma,Manan Mehta
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 9 Pages

点击查看摘要

[AI-149] he Hidden Game Problem

链接: https://arxiv.org/abs/2510.03845
作者: Gon Buzaglo,Noah Golowich,Elad Hazan
机构: 未知
类目: Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT); Machine Learning (cs.LG); Machine Learning (stat.ML)
备注:

点击查看摘要

[AI-150] A4FN: an Agent ic AI Architecture for Autonomous Flying Networks

链接: https://arxiv.org/abs/2510.03829
作者: André Coelho,Pedro Ribeiro,Helder Fontes,Rui Campos
机构: 未知
类目: Networking and Internet Architecture (cs.NI); Artificial Intelligence (cs.AI)
备注: This paper has been accepted for presentation in the Auto ML for Zero-Touch Network Management Workshop (WS04-01) at the IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC) 2025

点击查看摘要

[AI-151] Proximal Diffusion Neural Sampler

链接: https://arxiv.org/abs/2510.03824
作者: Wei Guo,Jaemoo Choi,Yuchen Zhu,Molei Tao,Yongxin Chen
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注: 31 pages, 12 figures

点击查看摘要

[AI-152] Detecting Invariant Manifolds in ReLU-Based RNNs

链接: https://arxiv.org/abs/2510.03814
作者: Lukas Eisenmann,Alena Brändle,Zahra Monfared,Daniel Durstewitz
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Dynamical Systems (math.DS)
备注:

点击查看摘要

[AI-153] 6G-Enabled Digital Twin Framework for Real-Time Cyber-Physical Systems: An Experimental Validation with Industrial Bearing Fault Detection

链接: https://arxiv.org/abs/2510.03807
作者: Vaskar Chakma,Wooyeol Choi
机构: 未知
类目: Networking and Internet Architecture (cs.NI); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[AI-154] Lightweight and Data-Efficient MultivariateTime Series Forecasting using Residual-Stacked Gaussian (RS-GLinear) Architecture

链接: https://arxiv.org/abs/2510.03788
作者: Abukar Ali
机构: 未知
类目: Computational Engineering, Finance, and Science (cs.CE); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-155] GuidedSampling: Steering LLM s Towards Diverse Candidate Solutions at Inference-Time

链接: https://arxiv.org/abs/2510.03777
作者: Divij Handa,Mihir Parmar,Aswin RRV,Md Nayem Uddin,Hamid Palangi,Chitta Baral
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[AI-156] OptAgent : Optimizing Query Rewriting for E-commerce via Multi-Agent Simulation

链接: https://arxiv.org/abs/2510.03771
作者: Divij Handa,David Blincoe,Orson Adams,Yinlin Fu
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-157] You Have Been LaTeXpOsEd: A Systematic Analysis of Information Leakage in Preprint Archives Using Large Language Models

链接: https://arxiv.org/abs/2510.03761
作者: Richard A. Dubniczky,Bertalan Borsos,Tihanyi Norbert
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-158] EvoEngineer: Mastering Automated CUDA Kernel Code Evolution with Large Language Models ICLR2026

链接: https://arxiv.org/abs/2510.03760
作者: Ping Guo,Chenyu Zhu,Siyuan Chen,Fei Liu,Xi Lin,Zhichao Lu,Qingfu Zhang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Under Review of ICLR 2026

点击查看摘要

[AI-159] Code4MeV2: a Research-oriented Code-completion Platform

链接: https://arxiv.org/abs/2510.03755
作者: Roham Koohestani,Parham Bateni,Aydin Ebrahimi,Behdad Etezadi,Kiarash Karimi,Maliheh Izadi
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: Under review for submission at a conference

点击查看摘要

[AI-160] HydroFusion-LMF: Semi-Supervised Multi-Network Fusion with Large-Model Adaptation for Long-Term Daily Runoff Forecasting

链接: https://arxiv.org/abs/2510.03744
作者: Qianfei Fan,Jiayu Wei,Peijun Zhu,Wensheng Ye,Meie Fang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC); Neural and Evolutionary Computing (cs.NE); Geophysics (physics.geo-ph)
备注: V1

点击查看摘要

[AI-161] Cost Efficient Fairness Audit Under Partial Feedback NEURIPS2025

链接: https://arxiv.org/abs/2510.03734
作者: Nirjhar Das,Mohit Sharma,Praharsh Nanavati,Kirankumar Shiragur,Amit Deshpande
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Machine Learning (stat.ML)
备注: Accepted at NeurIPS 2025 RegML Workshop; Reliable ML Workshop

点击查看摘要

[AI-162] H-DDx: A Hierarchical Evaluation Framework for Differential Diagnosis ALT NEURIPS2025

链接: https://arxiv.org/abs/2510.03700
作者: Seungseop Lim,Gibaeg Kim,Hyunkyung Lee,Wooseok Han,Jean Seo,Jaehyo Yoo,Eunho Yang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: GenAI4Health @NeurIPS 2025

点击查看摘要

[AI-163] REG: A Regularization Optimizer for Robust Training Dynamics

链接: https://arxiv.org/abs/2510.03691
作者: Zehua Liu,Han Wu,Xiaojin Fu,Shuqi Liu,Xiongwei Han,Tao Zhong,Mingxuan Yuan
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-164] Rainbow Padding: Mitigating Early Termination in Instruction-Tuned Diffusion LLM s

链接: https://arxiv.org/abs/2510.03680
作者: Bumjun Kim,Dongjae Jeon,Dueun Kim,Wonje Jeung,Albert No
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 25 pages. Project page available at~\url{ this https URL }

点击查看摘要

[AI-165] Operationalizing Data Minimization for Privacy-Preserving LLM Prompting

链接: https://arxiv.org/abs/2510.03662
作者: Jijie Zhou,Niloofar Mireshghallah,Tianshi Li
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-166] LLM -Guided Evolutionary Program Synthesis for Quasi-Monte Carlo Design

链接: https://arxiv.org/abs/2510.03650
作者: Amir Sadikov
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE); Neural and Evolutionary Computing (cs.NE); Numerical Analysis (math.NA)
备注:

点击查看摘要

[AI-167] Implicit Models: Expressive Power Scales with Test-Time Compute

链接: https://arxiv.org/abs/2510.03638
作者: Jialin Liu,Lisang Ding,Stanley Osher,Wotao Yin
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Representation Theory (math.RT); Machine Learning (stat.ML)
备注:

点击查看摘要

[AI-168] Predicting Stock Price Movement with LLM -Enhanced Tweet Emotion Analysis

链接: https://arxiv.org/abs/2510.03633
作者: An Vuong,Susan Gauch
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 17th International Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (KDIR 2025), Marbella, Spain, Oct. 22-24, 2025 (to appear) Best Student Paper Finalist

点击查看摘要

[AI-169] MITS: Enhanced Tree Search Reasoning for LLM s via Pointwise Mutual Information

链接: https://arxiv.org/abs/2510.03632
作者: Jiaxi Li,Yucheng Shi,Jin Lu,Ninghao Liu
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 18 pages

点击查看摘要

[AI-170] Explainable but Vulnerable: Adversarial Attacks on XAI Explanation in Cybersecurity Applications

链接: https://arxiv.org/abs/2510.03623
作者: Maraz Mia,Mir Mehedi A. Pritom
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 10 pages, 9 figures, 4 tables

点击查看摘要

[AI-171] Neural Bayesian Filtering

链接: https://arxiv.org/abs/2510.03614
作者: Christopher Solinas,Radovan Haluska,David Sychrovsky,Finbarr Timbers,Nolan Bard,Michael Buro,Martin Schmid,Nathan R. Sturtevant,Michael Bowling
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:

点击查看摘要

[AI-172] Cross-Modal Content Optimization for Steering Web Agent Preferences

链接: https://arxiv.org/abs/2510.03612
作者: Tanqiu Jiang,Min Bai,Nikolaos Pappas,Yanjun Qi,Sandesh Swamy
机构: 未知
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:

点击查看摘要

[AI-173] PentestMCP: A Toolkit for Agent ic Penetration Testing

【速读】：该论文旨在解决传统渗透测试（Penetration Testing）中高度依赖人工操作、效率低下且难以规模化的问题。解决方案的关键在于提出PentestMCP，一个基于Model-Context-Protocol架构的MCP服务器库，通过远程过程调用（RPC）范式实现多智能体（Multi-Agent）渗透测试流程的灵活构建与组合，从而支持网络扫描、服务指纹识别、漏洞探测、利用及后渗透等常见任务的自动化编排。

链接: https://arxiv.org/abs/2510.03610
作者: Zachary Ezetta,Wu-chang Feng
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

Abstract:Agentic AI is transforming security by automating many tasks being performed manually. While initial agentic approaches employed a monolithic architecture, the Model-Context-Protocol has now enabled a remote-procedure call (RPC) paradigm to agentic applications, allowing for the flexible construction and composition of multi-function agents. This paper describes PentestMCP, a library of MCP server implementations that support agentic penetration testing. By supporting common penetration testing tasks such as network scanning, resource enumeration, service fingerprinting, vulnerability scanning, exploitation, and post-exploitation, PentestMCP allows a developer to customize multi-agent workflows for performing penetration tests.
zh

[AI-174] Understanding the Role of Training Data in Test-Time Scaling

链接: https://arxiv.org/abs/2510.03605
作者: Adel Javanmard,Baharan Mirzasoleiman,Vahab Mirrokni
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Machine Learning (stat.ML)
备注: 24 pages, 4 figures

点击查看摘要

[AI-175] Deep Domain Adaptation for Turbofan Engine Remaining Useful Life Prediction: Methodologies Evaluation and Future Trends

链接: https://arxiv.org/abs/2510.03604
作者: Yucheng Wang,Mohamed Ragab,Yubo Hou,Zhenghua Chen,Min Wu,Xiaoli Li
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-176] Neon: Negative Extrapolation From Self-Training Improves Image Generation

链接: https://arxiv.org/abs/2510.03597
作者: Sina Alemohammad,Zhangyang Wang,Richard G. Baraniuk
机构: 未知
类目: Graphics (cs.GR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[AI-177] Deep Reinforcement Learning for Multi-Agent Coordination

链接: https://arxiv.org/abs/2510.03592
作者: Kehinde O. Aina,Sehoon Ha
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA); Robotics (cs.RO)
备注: 11 pages, 8 figures, 1 table, presented at SWARM 2022, to be published in Journal of Artificial Life and Robotics

点击查看摘要

[AI-178] Latent Mixture of Symmetries for Sample-Efficient Dynamic Learning

链接: https://arxiv.org/abs/2510.03578
作者: Haoran Li,Chenhan Xiao,Muhao Guo,Yang Weng
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注: 30 pages, 6 figures

点击查看摘要

[AI-179] Generalization of Graph Neural Network Models for Distribution Grid Fault Detection

链接: https://arxiv.org/abs/2510.03571
作者: Burak Karabulut,Carlo Manna,Chris Develder
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Systems and Control (eess.SY)
备注: This paper has been submitted and accepted for IEEE SmartGridComm 2025

点击查看摘要

[AI-180] Certifiable Safe RLHF: Fixed-Penalty Constraint Optimization for Safer Language Models

链接: https://arxiv.org/abs/2510.03520
作者: Kartik Pandit,Sourav Ganguly,Arnesh Banerjee,Shaahin Angizi,Arnob Ghosh
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Systems and Control (eess.SY)
备注:

点击查看摘要

[AI-181] OneFlow: Concurrent Mixed-Modal and Interleaved Generation with Edit Flows

链接: https://arxiv.org/abs/2510.03506
作者: John Nguyen,Marton Havasi,Tariq Berrada,Luke Zettlemoyer,Ricky T. Q. Chen
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: this https URL

点击查看摘要

[AI-182] Agent Hub: A Research Agenda for Agent Sharing Infrastructure

链接: https://arxiv.org/abs/2510.03495
作者: Erik Pautsch,Tanmay Singla,Wenxin Jiang,Huiyun Peng,Behnaz Hassanshahi,Konstantin Läufer,George K.Thiruvathukal,James C. Davis
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-183] Reasoning -based Anomaly Detection Framework: A Real-time Scalable and Automated Approach to Anomaly Detection Across Domains

链接: https://arxiv.org/abs/2510.03486
作者: Anupam Panwar,Himadri Pal,Jiali Chen,Kyle Cho,Riddick Jiang,Miao Zhao,Rajiv Krishnamurthy
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 11 pages, 7 figures

点击查看摘要

[AI-184] owards Policy-Compliant Agents : Learning Efficient Guardrails For Policy Violation Detection

链接: https://arxiv.org/abs/2510.03485
作者: Xiaofei Wen,Wenjie Jacky Mo,Yanan Xie,Peng Qi,Muhao Chen
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 16 pages, 5 figures

点击查看摘要

[AI-185] Destination-to-Chutes Task Mapping Optimization for Multi-Robot Coordination in Robotic Sorting Systems

链接: https://arxiv.org/abs/2510.03472
作者: Yulun Zhang,Alexandre O. G. Barbosa,Federico Pecora,Jiaoyang Li
机构: 未知
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: Accepted to IEEE International Symposium on Multi-Robot and Multi-Agent Systems (MRS) 2025

点击查看摘要

[AI-186] Bridging LLM Planning Agents and Formal Methods: A Case Study in Plan Verification

链接: https://arxiv.org/abs/2510.03469
作者: Keshav Ramani,Vali Tawosi,Salwa Alamir,Daniel Borrajo
机构: 未知
类目: Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO)
备注:

点击查看摘要

[AI-187] ALMAS: an Autonomous LLM -based Multi-Agent Software Engineering Framework

链接: https://arxiv.org/abs/2510.03463
作者: Vali Tawosi,Keshav Ramani,Salwa Alamir,Xiaomo Liu
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-188] A Qualitative Comparative Evaluation of Cognitive and Generative Theories

链接: https://arxiv.org/abs/2510.03453
作者: Paul S. Rosenbloom
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: To appear in Proceedings of the 12th Annual Conference on Advances in Cognitive Systems (ACS-25)

点击查看摘要

[AI-189] he Argument is the Explanation: Structured Argumentation for Trust in Agents

链接: https://arxiv.org/abs/2510.03442
作者: Ege Cakar,Per Ola Kristensson
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 8 pages, 4 figures, 6 tables, submitted to IAAI-26

点击查看摘要

[AI-190] Scalable Ground Station Selection for Large LEO Constellations

链接: https://arxiv.org/abs/2510.03438
作者: Grace Ra Kim,Duncan Eddy,Vedant Srinivas,Mykel J. Kochenderfer
机构: 未知
类目: Networking and Internet Architecture (cs.NI); Artificial Intelligence (cs.AI); Systems and Control (eess.SY)
备注: 14 pages, 7 tables, 10 figures, submitted to IEEE Aeroconf 2026

点击查看摘要

[AI-191] Generalized Orders of Magnitude for Scalable Parallel High-Dynamic-Range Computation

链接: https://arxiv.org/abs/2510.03426
作者: Franz A. Heinsen,Leo Kozachkov
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Numerical Analysis (math.NA)
备注: 18 pages, 4 figures (main text). 14 pages, 21 figures (appendix)

点击查看摘要

[AI-192] Multi-task neural diffusion processes for uncertainty-quantified wind power prediction

链接: https://arxiv.org/abs/2510.03419
作者: Joseph Rawson,Domniki Ladopoulou,Petros Dellaportas
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Applications (stat.AP); Machine Learning (stat.ML)
备注: 36 pages, 13 figures, 2 tables,

点击查看摘要

[AI-193] ContraG en: A Multi-Agent Generation Framework for Enterprise Contradictions Detection

链接: https://arxiv.org/abs/2510.03418
作者: Ananya Mantravadi,Shivali Dalmia,Abhishek Mukherji,Nand Dave,Anudha Mittal
机构: 未知
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:

点击查看摘要

[AI-194] NEXUS: Network Exploration for eXploiting Unsafe Sequences in Multi-Turn LLM Jailbreaks

链接: https://arxiv.org/abs/2510.03417
作者: Javad Rafiei Asl,Sidhant Narula,Mohammad Ghasemigol,Eduardo Blanco,Daniel Takabi
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: Javad Rafiei Asl and Sidhant Narula are co-first authors

点击查看摘要

[AI-195] Report of the 2025 Workshop on Next-Generation Ecosystems for Scientific Computing: Harnessing Community Software and AI for Cross-Disciplinary Team Science

链接: https://arxiv.org/abs/2510.03413
作者: L.C. McInnes,D. Arnold,P. Balaprakash,M. Bernhardt,B. Cerny,A. Dubey,R. Giles,D.W. Hood,M.A. Leung,V. Lopez-Marrero,P. Messina,O.B. Newton,C. Oehmen,S.M. Wild,J. Willenbring,L. Woodley,T. Baylis,D.E. Bernholdt,C. Camano,J. Cohoon,C. Ferenbaugh,S.M. Fiore,S. Gesing,D. Gomez-Zara,J. Howison,T. Islam,D. Kepczynski,C. Lively,H. Menon,B. Messer,M. Ngom,U. Paliath,M.E. Papka,I. Qualters,E.M. Raybourn,K. Riley,P. Rodriguez,D. Rouson,M. Schwalbe,S.K. Seal,O. Surer,V. Taylor,L. Wu
机构: 未知
类目: Computational Engineering, Finance, and Science (cs.CE); Artificial Intelligence (cs.AI); Mathematical Software (cs.MS)
备注: 38 pages, 6 figures

点击查看摘要

[AI-196] LegalSim: Multi-Agent Simulation of Legal Systems for Discovering Procedural Exploits EMNLP2025

链接: https://arxiv.org/abs/2510.03405
作者: Sanket Badhe
机构: 未知
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注: 12 pages with 2 figures, accepted at the NLLP workshop at EMNLP 2025

点击查看摘要

[AI-197] Cross-Modal Reconstruction Pretraining for Ramp Flow Prediction at Highway Interchanges

链接: https://arxiv.org/abs/2510.03381
作者: Yongchao Li,Jun Chen,Zhuoxuan Li,Chao Gao,Yang Li,Chu Zhang,Changyin Dong
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-198] A Robust Clustered Federated Learning Approach for Non-IID Data with Quantity Skew

链接: https://arxiv.org/abs/2510.03380
作者: Michael Ben Ali(IRIT, IRIT-SIG, UT3),Imen Megdiche(IRIT, IRIT-SIG, INUC),André Peninou(IRIT, IRIT-SIG, UT2J),Olivier Teste(IRIT-SIG, IRIT, UT2J, Comue de Toulouse)
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-199] Can an AI-Powered Presentation Platform Based On The Game “Just a Minute” Be Used To Improve Students Public Speaking Skills?

链接: https://arxiv.org/abs/2510.03379
作者: Frederic Higham,Tommy Yuan
机构: 未知
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注: 11 pages, to be presented orally at the International Conference on Education and Artificial Intelligence Technologies (Nov 2025)

点击查看摘要

[AI-200] Refined Iterated Pareto Greedy for Energy-aware Hybrid Flowshop Scheduling with Blocking Constraints

链接: https://arxiv.org/abs/2510.03377
作者: Ahmed Missaoui,Cemalettin Ozturk,Barry O’Sullivan
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-201] Distributed Low-Communication Training with Decoupled Momentum Optimization NEURIPS2025

链接: https://arxiv.org/abs/2510.03371
作者: Sasho Nedelkoski,Alexander Acker,Odej Kao,Soeren Becker,Dominik Scheinert
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC)
备注: NeurIPS 2025 - DynaFront 2025: Dynamics at the Frontiers of Optimization, Sampling, and Games Workshop

点击查看摘要

[AI-202] riQuest:An AI Copilot-Powered Platform for Interdisciplinary Curriculum Design

链接: https://arxiv.org/abs/2510.03369
作者: Huazhen Wang,Huimin Yang,Hainbin Lin,Yan Dong,Lili Chen,Liangliang Xia,Wenwen Xu
机构: 未知
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注: 16 pages, 4 figures

点击查看摘要

[AI-203] An Adaptive Responsible AI Governance Framework for Decentralized Organizations

链接: https://arxiv.org/abs/2510.03368
作者: Kiana Jafari Meimandi,Anka Reuel,Gabriela Aranguiz-Dias,Hatim Rahama,Ala-Eddine Ayadi,Xavier Boullier,Jérémy Verdo,Louis Montanie,Mykel Kochenderfer
机构: 未知
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-204] Disentangling Recall and Reasoning in Transformer Models through Layer-wise Attention and Activation Analysis

链接: https://arxiv.org/abs/2510.03366
作者: Harshwardhan Fartale,Ashish Kattamuri,Rahul Raja,Arpita Vats,Ishita Prasad,Akshata Kishore Moharir
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-205] Diffusion-Based Data-Assimilation-Enabled Super-Resolution of Hub-height Winds

链接: https://arxiv.org/abs/2510.03364
作者: Xiaolong Ma,Xu Dong,Ashley Tarrant,Lei Yang,Rao Kotamarthi,Jiali Wang,Feng Yan,Rajkumar Kettimuthu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-206] Physics-informed Neural-operator Predictive Control for Drag Reduction in Turbulent Flows

链接: https://arxiv.org/abs/2510.03360
作者: Zelin Zhao,Zongyi Li,Kimia Hassibi,Kamyar Azizzadenesheli,Junchi Yan,H. Jane Bae,Di Zhou,Anima Anandkumar
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Optimization and Control (math.OC); Fluid Dynamics (physics.flu-dyn)
备注:

点击查看摘要

[AI-207] Understanding Transformers for Time Series: Rank Structure Flow-of-ranks and Compressibility

链接: https://arxiv.org/abs/2510.03358
作者: Annan Yu,Danielle C. Maddix,Boran Han,Xiyuan Zhang,Abdul Fatir Ansari,Oleksandr Shchur,Christos Faloutsos,Andrew Gordon Wilson,Michael W. Mahoney,Yuyang Wang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 42 pages

点击查看摘要

[AI-208] Interpretable Neuropsychiatric Diagnosis via Concept-Guided Graph Neural Networks

链接: https://arxiv.org/abs/2510.03351
作者: Song Wang,Zhenyu Lei,Zhen Tan,Jundong Li,Javier Rasero,Aiying Zhang,Chirag Agarwal
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Image and Video Processing (eess.IV)
备注:

点击查看摘要

[AI-209] KVComm: Enabling Efficient LLM Communication through Selective KV Sharing

链接: https://arxiv.org/abs/2510.03346
作者: Xiangyu Shi,Marco Chiesa,Gerald Q. Maguire Jr.,Dejan Kostic
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:

点击查看摘要

[AI-210] Pilot selection in the era of Virtual reality: algorithms for accurate and interpretable machine learning models

链接: https://arxiv.org/abs/2510.03345
作者: Luoma Ke,Guangpeng Zhang,Jibo He,Yajing Li,Yan Li,Xufeng Liu,Peng Fang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-211] Defining a Strategic Action Plan for AI in Higher Education

链接: https://arxiv.org/abs/2510.03343
作者: Nikolaos Avouris
机构: 未知
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注: to be cited: N. Avouris (2025), Defining a Strategic Action Plan for AI in Higher Education, Proceedings International Scientific Conference on Digital Competencies in Higher Education, Tirana, September 2025, pp. 141-151

点击查看摘要

[AI-212] Learning Pareto-Optimal Pandemic Intervention Policies with MORL

链接: https://arxiv.org/abs/2510.03340
作者: Marian Chen,Miri Zilka
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Populations and Evolution (q-bio.PE)
备注:

点击查看摘要

[AI-213] Pool Me Wisely: On the Effect of Pooling in Transformer-Based Models

链接: https://arxiv.org/abs/2510.03339
作者: Sofiane Ennadir,Levente Zólyomi,Oleg Smirnov,Tianze Wang,John Pertoft,Filip Cornell,Lele Cao
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-214] Linguistic and Audio Embedding-Based Machine Learning for Alzheimers Dementia and Mild Cognitive Impairment Detection: Insights from the PROCESS Challenge

链接: https://arxiv.org/abs/2510.03336
作者: Adharsha Sam Edwin Sam Devahi,Sohail Singh Sangha,Prachee Priyadarshinee,Jithin Thilakan,Ivan Fu Xing Tan,Christopher Johann Clarke,Sou Ka Lon,Balamurali B T,Yow Wei Quin,Chen Jer-Ming
机构: 未知
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[AI-215] Intelligent Healthcare Ecosystems: Optimizing the Iron Triangle of Healthcare (Access Cost Quality)

链接: https://arxiv.org/abs/2510.03331
作者: Vivek Acharya
机构: 未知
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注: 8 pages, 4 figures, formatted per MDPI guidelines, APA-style numbered references

点击查看摘要

[AI-216] Predicting Effects Missing Distributions: Evaluating LLM s as Human Behavior Simulators in Operations Management

链接: https://arxiv.org/abs/2510.03310
作者: Runze Zhang,Xiaowei Zhang,Mingyang Zhao
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-217] Dynamic Meta-Learning for Adaptive XGBoost-Neural Ensembles

链接: https://arxiv.org/abs/2510.03301
作者: Arthur Sedek
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-218] From Score Distributions to Balance: Plug-and-Play Mixture-of-Experts Routing

链接: https://arxiv.org/abs/2510.03293
作者: Rana Shahout,Colin Cai,Yilun Du,Minlan Yu,Michael Mitzenmacher
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC)
备注:

点击查看摘要

[AI-219] UniPruning: Unifying Local Metric and Global Feedback for Scalable Sparse LLM s

链接: https://arxiv.org/abs/2510.03291
作者: Yizhuo Ding,Wanying Qu,Jiawei Geng,Wenqi Shao,Yanwei Fu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-220] LogAction: Consistent Cross-system Anomaly Detection through Logs via Active Domain

链接: https://arxiv.org/abs/2510.03288
作者: Chiming Duan,Minghua He,Pei Xiao,Tong Jia,Xin Zhang,Zhewei Zhong,Xiang Luo,Yan Niu,Lingzhe Zhang,Yifan Wu,Siyu Yu,Weijie Hong,Ying Li,Gang Huang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC); Software Engineering (cs.SE)
备注: The 40th IEEE/ACM International Conference on Automated Software Engineering, ASE 2025

点击查看摘要

[AI-221] WAREX: Web Agent Reliability Evaluation on Existing Benchmarks

链接: https://arxiv.org/abs/2510.03285
作者: Su Kara,Fazle Faisal,Suman Nath
机构: 未知
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
备注:

点击查看摘要

[AI-222] Edge-FIT: Federated Instruction Tuning of Quantized LLM s for Privacy-Preserving Smart Home Environments

链接: https://arxiv.org/abs/2510.03284
作者: Vinay Venkatesh,Vamsidhar R Kamanuru,Lav Kumar,Nikita Kothari
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 7 pages, 1 figure

点击查看摘要

[AI-223] Quantifying constraint hierarchies in Bayesian PINNs via per-constraint Hessian decomposition

链接: https://arxiv.org/abs/2510.03278
作者: Filip Landgren
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 5 pages, 2 figures

点击查看摘要

[AI-224] QuadEnhancer: Leverag ing Quadratic Transformations to Enhance Deep Neural Networks NEURIPS2025

链接: https://arxiv.org/abs/2510.03276
作者: Qian Chen,Linxin Yang,Akang Wang,Xiaodong Luo,Yin Zhang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 39th Conference on Neural Information Processing Systems (NeurIPS 2025)

点击查看摘要

[AI-225] Quant-dLLM : Post-Training Extreme Low-Bit Quantization for Diffusion Large Language Models

链接: https://arxiv.org/abs/2510.03274
作者: Tianao Zhang,Zhiteng Li,Xianglong Yan,Haotong Qin,Yong Guo,Yulun Zhang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-226] Learning without Global Backpropagation via Synergistic Information Distillation

链接: https://arxiv.org/abs/2510.03273
作者: Chenhao Ye,Ming Tang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-227] PDE-Transformer: A Continuous Dynamical Systems Approach to Sequence Modeling

链接: https://arxiv.org/abs/2510.03272
作者: Yukun Zhang,Xueqing Zhou
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-228] Decision Potential Surface: A Theoretical and Practical Approximation of LLM s Decision Boundary

链接: https://arxiv.org/abs/2510.03271
作者: Zi Liang,Zhiyao Wu,Haoyang Shang,Yulin Jin,Qingqing Ye,Huadi Zheng,Peizhao Hu,Haibo Hu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Source code: this https URL

点击查看摘要

[AI-229] CoDA: Coding LM via Diffusion Adaptation

链接: https://arxiv.org/abs/2510.03270
作者: Haolin Chen,Shiyu Wang,Can Qin,Bo Pang,Zuxin Liu,Jielin Qiu,Jianguo Zhang,Yingbo Zhou,Zeyuan Chen,Ran Xu,Shelby Heinecke,Silvio Savarese,Caiming Xiong,Huan Wang,Weiran Yao
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-230] Decrypt Modality Gap in Multimodal Contrastive Learning: From Convergent Representation to Pair Alignment

链接: https://arxiv.org/abs/2510.03268
作者: Lingjie Yi,Raphael Douady,Chao Chen
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-231] PT2-LLM : Post-Training Ternarization for Large Language Models

链接: https://arxiv.org/abs/2510.03267
作者: Xianglong Yan,Chengzhu Bao,Zhiteng Li,Tianao Zhang,Kaicheng Yang,Haotong Qin,Ruobing Xie,Xingwu Sun,Yulun Zhang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-232] MindCraft: How Concept Trees Take Shape In Deep Models

链接: https://arxiv.org/abs/2510.03265
作者: Bowei Tian,Yexiao He,Wanghao Ye,Ziyao Wang,Meng Liu,Ang Li
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-233] Front-Loading Reasoning : The Synergy between Pretraining and Post-Training Data

链接: https://arxiv.org/abs/2510.03264
作者: Syeda Nahida Akter,Shrimai Prabhumoye,Eric Nyberg,Mostofa Patwary,Mohammad Shoeybi,Yejin Choi,Bryan Catanzaro
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-234] Memory Self-Regeneration: Uncovering Hidden Knowledge in Unlearned Models

链接: https://arxiv.org/abs/2510.03263
作者: Agnieszka Polowczyk,Alicja Polowczyk,Joanna Waczyńska,Piotr Borycki,Przemysław Spurek
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-235] Semantic-Inductive Attribute Selection for Zero-Shot Learning

链接: https://arxiv.org/abs/2510.03260
作者: Juan Jose Herrera-Aranda,Guillermo Gomez-Trenado,Francisco Herrera,Isaac Triguero
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 26 pages, 9 figures, code available at this https URL

点击查看摘要

[AI-236] Meta-Awareness Enhances Reasoning Models: Self-Alignment Reinforcement Learning

链接: https://arxiv.org/abs/2510.03259
作者: Yoonjeon Kim,Doohyuk Jang,Eunho Yang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: preprint

点击查看摘要

[AI-237] POEM: Explore Unexplored Reliable Samples to Enhance Test-Time Adaptation

链接: https://arxiv.org/abs/2510.03258
作者: Chang’an Yi,Xiaohui Deng,Shuaicheng Niu,Yan Zhou
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 11pages,6 figures

点击查看摘要

[AI-238] riple-BERT: Do We Really Need MARL for Order Dispatch on Ride-Sharing Platforms?

链接: https://arxiv.org/abs/2510.03257
作者: Zijian Zhao,Sen Li
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:

点击查看摘要

[AI-239] SciTS: Scientific Time Series Understanding and Generation with LLM s

链接: https://arxiv.org/abs/2510.03255
作者: Wen Wu,Ziyang Zhang,Liwei Liu,Xuenan Xu,Junlin Liu,Ke Fan,Qitan Lv,Jimin Zhuang,Chen Zhang,Zheqi Yuan,Siyuan Hou,Tianyi Lin,Kai Chen,Bowen Zhou,Chao Zhang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-240] Solving the Granularity Mismatch: Hierarchical Preference Learning for Long-Horizon LLM Agents

链接: https://arxiv.org/abs/2510.03253
作者: Heyang Gao,Zexu Sun,Erxue Min,Hengyi Cai,Shuaiqiang Wang,Dawei Yin,Xu Chen
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Preprint

点击查看摘要

[AI-241] Numerion: A Multi-Hypercomplex Model for Time Series Forecasting

链接: https://arxiv.org/abs/2510.03251
作者: Hanzhong Cao,Wenbo Yan,Ying Tan
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-242] owards Multimodal Active Learning: Efficient Learning with Limited Paired Data

链接: https://arxiv.org/abs/2510.03247
作者: Jiancheng Zhang,Yinglun Zhu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-243] StructPrune: Structured Global Pruning asymptotics with mathcalO(sqrtN) GPU Memory

【速读】：该论文旨在解决大规模语言模型（Large Language Models, LLMs）在结构化剪枝（Structured Pruning）过程中面临的内存消耗过高与性能下降的矛盾问题。传统全局剪枝虽能实现优异性能，但其内存复杂度为 $\mathcal{O}(N)$ ，难以应用于百亿参数级别的模型；而局部剪枝虽降低内存至单层水平，却因忽略层间依赖关系，在高稀疏率下表现不佳。解决方案的关键在于提出一种基于交替方向乘子法（ADMM）的分治策略——STRUPRUNE，它将全局剪枝问题分解为多个可并行处理的子问题，使每个模块均能在有限GPU内存内完成优化，同时通过推导出结构化剪枝掩码的闭式解析解和基于能量的渐近分配框架，实现了层间稀疏度的显式分配与自适应调整，从而在保持硬件友好性的同时显著降低内存需求至 $\mathcal{O}(\sqrt{N})$ ，兼顾性能与可扩展性。

链接: https://arxiv.org/abs/2510.03246
作者: Xinyuan Song,Guangji Bai,Liang Zhao
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

Abstract:Pruning is critical for scaling large language models (LLMs). Global pruning achieves strong performance but requires \mathcalO(N) memory, which is infeasible for billion-parameter models. Local pruning reduces GPU memory usage to that of a single layer by pruning layers independently, but it neglects inter-layer dependencies and often leads to suboptimal performance in high-sparsity regimes. Unlike unstructured pruning, structured pruning produces regular sparsity patterns that align well with GPU kernels and library optimizations, making it more hardware-efficient. However, structured pruning typically relies on global pruning, since structured patterns are more prone to severe performance degradation under local optimization. To jointly achieve structured pruning and the memory efficiency of local pruning, we propose a divide-and-conquer strategy that decomposes the global pruning problem into coordinated subproblems across different modules, each of which fits within limited GPU memory. Building on this idea, we design \textbfSTRUPRUNE, an ADMM-based framework that integrates structured sparsity into the pruning process, combining the memory efficiency of local pruning with the hardware compatibility of structured methods. We derive a closed-form analytical solution for structured pruning masks that provides an explicit rule for layer-wise sparsity allocation, and further develop an energy-based asymptotic framework yielding a softmax-form allocation scheme that simplifies optimization while adapting to heterogeneous layer importance. Experiments demonstrate that STRUPRUNE matches the perplexity of global structured pruning while reducing memory cost from \mathcalO(N) to \mathcalO(\sqrtN) , enabling practical deployment at the billion-parameter scale.
zh

[AI-244] PARS: Low-Latency LLM Serving via Pairwise Learning-to-Rank

链接: https://arxiv.org/abs/2510.03243
作者: Yiheng Tao,Yihe Zhang,Matthew T. Dearing,Xin Wang,Yuping Fan,Zhiling Lan
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC); Performance (cs.PF)
备注:

点击查看摘要

[AI-245] Neural Network Surrogates for Free Energy Computation of Complex Chemical Systems ICSE

链接: https://arxiv.org/abs/2510.01396
作者: Wasut Pornpatcharapong
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE); Chemical Physics (physics.chem-ph); Computational Physics (physics.comp-ph)
备注: 6 pages, 4 figures. This work has already been accepted for presentation in The 29th International Computer Science and Engineering Conference (ICSEC) 2025, Chiang Mai, Thailand, and will be published in IEEE Xplore

点击查看摘要

[AI-246] Embracing Discrete Search: A Reason able Approach to Causal Structure Learning

链接: https://arxiv.org/abs/2510.04970
作者: Marcel Wienöbst,Leonard Henckel,Sebastian Weichwald
机构: 未知
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Methodology (stat.ME)
备注:

点击查看摘要

[AI-247] MuFFIN: Multifaceted Pronunciation Feedback Model with Interactive Hierarchical Neural Modeling

链接: https://arxiv.org/abs/2510.04956
作者: Bi-Cheng Yan,Ming-Kang Tsai,Berlin Chen
机构: 未知
类目: Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI)
备注: Submitted to IEEE/ACM Transactions on Audio, Speech, and Language Processing

点击查看摘要

[AI-248] AURA Score: A Metric For Holistic Audio Question Answering Evaluation

链接: https://arxiv.org/abs/2510.04934
作者: Satvik Dixit,Soham Deshmukh,Bhiksha Raj
机构: 未知
类目: Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-249] Fisher-Bingham-like normalizing flows on the sphere

链接: https://arxiv.org/abs/2510.04762
作者: Thorsten Glüsenkamp
机构: 未知
类目: Machine Learning (stat.ML); Instrumentation and Methods for Astrophysics (astro-ph.IM); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[AI-250] he Bayesian Origin of the Probability Weighting Function in Human Representation of Probabilities

链接: https://arxiv.org/abs/2510.04698
作者: Xin Tong,Thi Thu Uyen Hoang,Xue-Xin Wei,Michael Hahn
机构: 未知
类目: Neurons and Cognition (q-bio.NC); Artificial Intelligence (cs.AI); Theoretical Economics (econ.TH)
备注:

点击查看摘要

[AI-251] Computing Wasserstein Barycenters through Gradient Flows

链接: https://arxiv.org/abs/2510.04602
作者: Eduardo Fernandes Montesuma,Yassir Bendou,Mike Gartrell
机构: 未知
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 4 Figures, 3 Tables, under review

点击查看摘要

[AI-252] Inverse Mixed-Integer Programming: Learning Constraints then Objective Functions

链接: https://arxiv.org/abs/2510.04455
作者: Akira Kitaoka
机构: 未知
类目: Optimization and Control (math.OC); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Statistics Theory (math.ST); Machine Learning (stat.ML)
备注: 33 pages

点击查看摘要

[AI-253] Scalable Causal Discovery from Recursive Nonlinear Data via Truncated Basis Function Scores and Tests

链接: https://arxiv.org/abs/2510.04276
作者: Joseph Ramsey,Bryan Andrews
机构: 未知
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI)
备注: 30 pages, 11 figures, 5 tables

点击查看摘要

[AI-254] A Contextual Quality Reward Model for Reliable and Efficient Best-of-N Sampling

链接: https://arxiv.org/abs/2510.04087
作者: Hyung Gyu Rho
机构: 未知
类目: Methodology (stat.ME); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:

点击查看摘要

[AI-255] ReTiDe: Real-Time Denoising for Energy-Efficient Motion Picture Processing with FPGAs SIGGRAPH

链接: https://arxiv.org/abs/2510.03812
作者: Changhong Li,Clément Bled,Rosa Fernandez,Shreejith Shanker
机构: 未知
类目: Image and Video Processing (eess.IV); Artificial Intelligence (cs.AI)
备注: This paper has been accepted by the 22nd ACM SIGGRAPH European Conference on Visual Media Production (CVMP 2025)

点击查看摘要

[AI-256] Dissecting Larval Zebrafish Hunting using Deep Reinforcement Learning Trained RNN Agents

链接: https://arxiv.org/abs/2510.03699
作者: Raaghav Malik,Satpreet H. Singh,Sonja Johnson-Yu,Nathan Wu,Roy Harpaz,Florian Engert,Kanaka Rajan
机构: 未知
类目: Neurons and Cognition (q-bio.NC); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE); Systems and Control (eess.SY)
备注:

点击查看摘要

[AI-257] Deep learning the sources of MJO predictability: a spectral view of learned features

链接: https://arxiv.org/abs/2510.03582
作者: Lin Yao,Da Yang,James P.C. Duncan,Ashesh Chattopadhyay,Pedram Hassanzadeh,Wahid Bhimji,Bin Yu
机构: 未知
类目: Atmospheric and Oceanic Physics (physics.ao-ph); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-258] Agile Tradespace Exploration for Space Rendezvous Mission Design via Transformers

链接: https://arxiv.org/abs/2510.03544
作者: Yuji Takubo,Daniele Gammelli,Marco Pavone,Simone D’Amico
机构: 未知
类目: Optimization and Control (math.OC); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注: 14 pages, 7 figures

点击查看摘要

[AI-259] Application of a Virtual Imaging Framework for Investigating a Deep Learning-Based Reconstruction Method for 3D Quantitative Photoacoustic Computed Tomography

链接: https://arxiv.org/abs/2510.03431
作者: Refik Mert Cam,Seonyeong Park,Umberto Villa,Mark A. Anastasio
机构: 未知
类目: Medical Physics (physics.med-ph); Artificial Intelligence (cs.AI); Signal Processing (eess.SP)
备注: Preprint submitted to Elsevier Photoacoustics

点击查看摘要

[AI-260] InstructPLM-mu: 1-Hour Fine-Tuning of ESM2 Beats ESM3 in Protein Mutation Predictions

链接: https://arxiv.org/abs/2510.03370
作者: Junde Xu,Yapin Shi,Lijun Lang,Taoyong Cui,Zhiming Zhang,Guangyong Chen,Jiezhong Qiu,Pheng-Ann Heng
机构: 未知
类目: Quantitative Methods (q-bio.QM); Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE)
备注: preprint

点击查看摘要

[AI-261] NS-Pep: De novo Peptide Design with Non-Standard Amino Acids

链接: https://arxiv.org/abs/2510.03326
作者: Tao Guo,Junbo Yin,Yu Wang,Xin Gao
机构: 未知
类目: Biomolecules (q-bio.BM); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

[AI-262] Atlas-free Brain Network Transformer

链接: https://arxiv.org/abs/2510.03306
作者: Shuai Huang,Xuan Kan,James J. Lah,Deqiang Qiu
机构: 未知
类目: Neurons and Cognition (q-bio.NC); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE); Image and Video Processing (eess.IV)
备注:

点击查看摘要

[AI-263] A Biologically Interpretable Cognitive Architecture for Online Structuring of Episodic Memories into Cognitive Maps

链接: https://arxiv.org/abs/2510.03286
作者: E.A. Dzhivelikian,A.I. Panov
机构: 未知
类目: Neurons and Cognition (q-bio.NC); Artificial Intelligence (cs.AI)
备注:

点击查看摘要

机器学习

[LG-0] MICROTRIPS: MICRO-geography TRavel Intelligence and Pattern Synthesis

链接: https://arxiv.org/abs/2510.05080
作者: Yangyang Wang,Tayo Fabusuyi
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-1] ResMimic: From General Motion Tracking to Humanoid Whole-body Loco-Manipulation via Residual Learning

链接: https://arxiv.org/abs/2510.05070
作者: Siheng Zhao,Yanjie Ze,Yue Wang,C. Karen Liu,Pieter Abbeel,Guanya Shi,Rocky Duan
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: 9 pages, 8 figures

点击查看摘要

Abstract:Humanoid whole-body loco-manipulation promises transformative capabilities for daily service and warehouse tasks. While recent advances in general motion tracking (GMT) have enabled humanoids to reproduce diverse human motions, these policies lack the precision and object awareness required for loco-manipulation. To this end, we introduce ResMimic, a two-stage residual learning framework for precise and expressive humanoid control from human motion data. First, a GMT policy, trained on large-scale human-only motion, serves as a task-agnostic base for generating human-like whole-body movements. An efficient but precise residual policy is then learned to refine the GMT outputs to improve locomotion and incorporate object interaction. To further facilitate efficient training, we design (i) a point-cloud-based object tracking reward for smoother optimization, (ii) a contact reward that encourages accurate humanoid body-object interactions, and (iii) a curriculum-based virtual object controller to stabilize early training. We evaluate ResMimic in both simulation and on a real Unitree G1 humanoid. Results show substantial gains in task success, training efficiency, and robustness over strong baselines. Videos are available at this https URL .

[LG-2] Boomerang Distillation Enables Zero-Shot Model Size Interpolation

链接: https://arxiv.org/abs/2510.05064
作者: Sara Kangaslahti,Nihal V. Nayak,Jonathan Geuter,Marco Fumero,Francesco Locatello,David Alvarez-Melis
类目: Machine Learning (cs.LG)
*备注: 10 pages, 7 figures in main text

点击查看摘要

[LG-3] ResCP: Reservoir Conformal Prediction for Time Series Forecasting

链接: https://arxiv.org/abs/2510.05060
作者: Roberto Neglia,Andrea Cini,Michael M. Bronstein,Filippo Maria Bianchi
类目: Machine Learning (cs.LG); Statistics Theory (math.ST); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-4] Modeling Student Learning with 3.8 Million Program Traces

链接: https://arxiv.org/abs/2510.05056
作者: Alexis Ross,Megha Srivastava,Jeremiah Blanchard,Jacob Andreas
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

Abstract:As programmers write code, they often edit and retry multiple times, creating rich “interaction traces” that reveal how they approach coding tasks and provide clues about their level of skill development. For novice programmers in particular, these traces reflect the diverse reasoning processes they employ to code, such as exploratory behavior to understand how a programming concept works, re-strategizing in response to bugs, and personalizing stylistic choices. In this work, we explore what can be learned from training language models on such reasoning traces: not just about code, but about coders, and particularly students learning to program. We introduce a dataset of over 3.8 million programming reasoning traces from users of Pencil Code, a free online educational platform used by students to learn simple programming concepts. Compared to models trained only on final programs or synthetically-generated traces, we find that models trained on real traces are stronger at modeling diverse student behavior. Through both behavioral and probing analyses, we also find that many properties of code traces, such as goal backtracking or number of comments, can be predicted from learned representations of the students who write them. Building on this result, we show that we can help students recover from mistakes by steering code generation models to identify a sequence of edits that will results in more correct code while remaining close to the original student’s style. Together, our results suggest that many properties of code are properties of individual students and that training on edit traces can lead to models that are more steerable, more predictive of student behavior while programming, and better at generating programs in their final states. Code and data is available at this https URL

[LG-5] KEEP: Integrating Medical Ontologies with Clinical Data for Robust Code Embeddings

链接: https://arxiv.org/abs/2510.05049
作者: Ahmed Elhussein,Paul Meddeb,Abigail Newbury,Jeanne Mirone,Martin Stoll,Gamze Gursoy
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

Abstract:Machine learning in healthcare requires effective representation of structured medical codes, but current methods face a trade off: knowledge graph based approaches capture formal relationships but miss real world patterns, while data driven methods learn empirical associations but often overlook structured knowledge in medical terminologies. We present KEEP (Knowledge preserving and Empirically refined Embedding Process), an efficient framework that bridges this gap by combining knowledge graph embeddings with adaptive learning from clinical data. KEEP first generates embeddings from knowledge graphs, then employs regularized training on patient records to adaptively integrate empirical patterns while preserving ontological relationships. Importantly, KEEP produces final embeddings without task specific auxiliary or end to end training enabling KEEP to support multiple downstream applications and model architectures. Evaluations on structured EHR from UK Biobank and MIMIC IV demonstrate that KEEP outperforms both traditional and Language Model based approaches in capturing semantic relationships and predicting clinical outcomes. Moreover, KEEP’s minimal computational requirements make it particularly suitable for resource constrained environments.

[LG-6] Inoculation Prompting: Instructing LLM s to misbehave at train-time improves test-time alignment

链接: https://arxiv.org/abs/2510.05024
作者: Nevan Wichers,Aram Ebtekar,Ariana Azarbal,Victor Gillioz,Christine Ye,Emil Ryd,Neil Rathi,Henry Sleight,Alex Mallen,Fabien Roger,Samuel Marks
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-7] Power Transform Revisited: Numerically Stable and Federated

链接: https://arxiv.org/abs/2510.04995
作者: Xuefeng Xu,Graham Cormode
类目: Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注: 25 pages

点击查看摘要

[LG-8] Adaptive Memory Momentum via a Model-Based Framework for Deep Learning Optimization

链接: https://arxiv.org/abs/2510.04988
作者: Kristi Topollai,Anna Choromanska
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-9] Federated Computation of ROC and PR Curves

链接: https://arxiv.org/abs/2510.04979
作者: Xuefeng Xu,Graham Cormode
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注: 23 pages

点击查看摘要

Abstract:Receiver Operating Characteristic (ROC) and Precision-Recall (PR) curves are fundamental tools for evaluating machine learning classifiers, offering detailed insights into the trade-offs between true positive rate vs. false positive rate (ROC) or precision vs. recall (PR). However, in Federated Learning (FL) scenarios, where data is distributed across multiple clients, computing these curves is challenging due to privacy and communication constraints. Specifically, the server cannot access raw prediction scores and class labels, which are used to compute the ROC and PR curves in a centralized setting. In this paper, we propose a novel method for approximating ROC and PR curves in a federated setting by estimating quantiles of the prediction score distribution under distributed differential privacy. We provide theoretical bounds on the Area Error (AE) between the true and estimated curves, demonstrating the trade-offs between approximation accuracy, privacy, and communication cost. Empirical results on real-world datasets demonstrate that our method achieves high approximation accuracy with minimal communication and strong privacy guarantees, making it practical for privacy-preserving model evaluation in federated systems.

[LG-10] StructuralDecompose: A Modular Framework for Robust Time Series Decomposition in R

链接: https://arxiv.org/abs/2510.04974
作者: Allen Daniel Sunny
类目: Machine Learning (cs.LG)
*备注: 8 pages, 4 figures. Part of the R package StructuralDecompose ( this https URL )

点击查看摘要

Abstract:We present StructuralDecompose, an R package for modular and interpretable time series decomposition. Unlike existing approaches that treat decomposition as a monolithic process, StructuralDecompose separates the analysis into distinct components: changepoint detection, anomaly detection, smoothing, and decomposition. This design provides flexibility and robust- ness, allowing users to tailor methods to specific time series characteristics. We demonstrate the package on simulated and real-world datasets, benchmark its performance against state-of-the- art tools such as Rbeast and autostsm, and discuss its role in interpretable machine learning workflows.

[LG-11] Egalitarian Gradient Descent: A Simple Approach to Accelerated Grokking

链接: https://arxiv.org/abs/2510.04930
作者: Ali Saheb Pasand,Elvis Dohmatob
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-12] How Different from the Past? Spatio-Temporal Time Series Forecasting with Self-Supervised Deviation Learning NEURIPS2025

链接: https://arxiv.org/abs/2510.04908
作者: Haotian Gao,Zheng Dong,Jiawei Yong,Shintaro Fukushima,Kenjiro Taura,Renhe Jiang
类目: Machine Learning (cs.LG)
*备注: Accepted at NeurIPS 2025

点击查看摘要

[LG-13] DP-HYPE: Distributed Differentially Private Hyperparameter Search

链接: https://arxiv.org/abs/2510.04902
作者: Johannes Liebenow,Thorsten Peinemann,Esfandiar Mohammadi
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-14] Benchmarking M-LTSF: Frequency and Noise-Based Evaluation of Multivariate Long Time Series Forecasting Models

链接: https://arxiv.org/abs/2510.04900
作者: Nick Janßen,Melanie Schaller,Bodo Rosenhahn
类目: Machine Learning (cs.LG); Systems and Control (eess.SY)
*备注: Number of pages: 13 Number of figures: 16 Number of Tables: 1 Submitted to: IEEE Transactions on Signal Processing

点击查看摘要

[LG-15] RL Is a Hammer and LLM s Are Nails: A Simple Reinforcement Learning Recipe for Strong Prompt Injection

链接: https://arxiv.org/abs/2510.04885
作者: Yuxin Wen,Arman Zharmagambetov,Ivan Evtimov,Narine Kokhlikyan,Tom Goldstein,Kamalika Chaudhuri,Chuan Guo
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-16] Flow-Matching Based Refiner for Molecular Conformer Generation

链接: https://arxiv.org/abs/2510.04878
作者: Xiangyang Xu,Hongyang Gao
类目: Machine Learning (cs.LG); Quantitative Methods (q-bio.QM)
*备注:

点击查看摘要

Abstract:Low-energy molecular conformers generation (MCG) is a foundational yet challenging problem in drug discovery. Denoising-based methods include diffusion and flow-matching methods that learn mappings from a simple base distribution to the molecular conformer distribution. However, these approaches often suffer from error accumulation during sampling, especially in the low SNR steps, which are hard to train. To address these challenges, we propose a flow-matching refiner for the MCG task. The proposed method initializes sampling from mixed-quality outputs produced by upstream denoising models and reschedules the noise scale to bypass the low-SNR phase, thereby improving sample quality. On the GEOM-QM9 and GEOM-Drugs benchmark datasets, the generator-refiner pipeline improves quality with fewer total denoising steps while preserving diversity.

[LG-17] A Clinical-grade Universal Foundation Model for Intraoperative Pathology

链接: https://arxiv.org/abs/2510.04861
作者: Zihan Zhao,Fengtao Zhou,Ronggang Li,Bing Chu,Xinke Zhang,Xueyi Zheng,Ke Zheng,Xiaobo Wen,Jiabo Ma,Yihui Wang,Jiewei Chen,Chengyou Zheng,Jiangyu Zhang,Yongqin Wen,Jiajia Meng,Ziqi Zeng,Xiaoqing Li,Jing Li,Dan Xie,Yaping Ye,Yu Wang,Hao Chen,Muyan Cai
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-18] Synthesising Counterfactual Explanations via Label-Conditional Gaussian Mixture Variational Autoencoders

链接: https://arxiv.org/abs/2510.04855
作者: Junqi Jiang,Francesco Leofante,Antonio Rago,Francesca Toni
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-19] On the Hardness of Learning Regular Expressions

链接: https://arxiv.org/abs/2510.04834
作者: Idan Attias,Lev Reyzin,Nathan Srebro,Gal Vardi
类目: Machine Learning (cs.LG); Computational Complexity (cs.CC)
*备注:

点击查看摘要

[LG-20] MetaMP: Seamless Metadata Enrichment and AI Application Framework for Enhanced Membrane Protein Visualization and Analysis

链接: https://arxiv.org/abs/2510.04776
作者: Ebenezer Awotoro,Chisom Ezekannagha,Florian Schwarz,Johannes Tauscher,Dominik Heider,Katharina Ladewig,Christel Le Bon,Karine Moncoq,Bruno Miroux,Georges Hattab
类目: Machine Learning (cs.LG); Databases (cs.DB)
*备注:

点击查看摘要

[LG-21] ParallelBench: Understanding the Trade-offs of Parallel Decoding in Diffusion LLM s

链接: https://arxiv.org/abs/2510.04767
作者: Wonjun Kang,Kevin Galim,Seunghyuk Oh,Minjae Lee,Yuchen Zeng,Shuibai Zhang,Coleman Hooper,Yuezhou Hu,Hyung Il Koo,Nam Ik Cho,Kangwook Lee
类目: Machine Learning (cs.LG)
*备注: Project Page: this https URL

点击查看摘要

[LG-22] Provable Affine Identifiability of Nonlinear CCA under Latent Distributional Priors

链接: https://arxiv.org/abs/2510.04758
作者: Zhiwei Han,Stefan Matthes,Hao Shen
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-23] EVaR-Optimal Arm Identification in Bandits

链接: https://arxiv.org/abs/2510.04728
作者: Mehrasa Ahmadipour,Aurélien Garivier
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-24] Directional Sheaf Hypergraph Networks: Unifying Learning on Directed and Undirected Hypergraphs

链接: https://arxiv.org/abs/2510.04727
作者: Emanuele Mule,Stefano Fiorini,Antonio Purificato,Federico Siciliano,Stefano Coniglio,Fabrizio Silvestri
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-25] ViTs: Teaching Machines to See Time Series Anomalies Like Human Experts

链接: https://arxiv.org/abs/2510.04710
作者: Zexin Wang,Changhua Pei,Yang Liu,Hengyue Jiang,Quan Zhou,Haotian Si,Hang Cui,Jianhui Li,Gaogang Xie,Jingjing Li,Dan Pei
类目: Machine Learning (cs.LG)
*备注: 13 pages

点击查看摘要

[LG-26] A Study on the Data Distribution Gap in Music Emotion Recognition

链接: https://arxiv.org/abs/2510.04688
作者: Joann Ching,Gerhard Widmer
类目: ound (cs.SD); Machine Learning (cs.LG)
*备注: Accepted at the 17th International Symposium on Computer Music Multidisciplinary Research (CMMR) 2025

点击查看摘要

[LG-27] Parameter-free Algorithms for the Stochastically Extended Adversarial Model NEURIPS2025

链接: https://arxiv.org/abs/2510.04685
作者: Shuche Wang,Adarsh Barik,Peng Zhao,Vincent Y. F. Tan
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注: Accepted to NeurIPS 2025

点击查看摘要

[LG-28] Counterfactual Credit Guided Bayesian Optimization

链接: https://arxiv.org/abs/2510.04676
作者: Qiyu Wei,Haowei Wang,Richard Allmendinger,Mauricio A. Álvarez
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-29] IMLP: An Energy-Efficient Continual Learning Method for Tabular Data Streams

链接: https://arxiv.org/abs/2510.04660
作者: Yuandou Wang,Filip Gunnarsson,Rihan Hai
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

Abstract:Tabular data streams are rapidly emerging as a dominant modality for real-time decision-making in healthcare, finance, and the Internet of Things (IoT). These applications commonly run on edge and mobile devices, where energy budgets, memory, and compute are strictly limited. Continual learning (CL) addresses such dynamics by training models sequentially on task streams while preserving prior knowledge and consolidating new knowledge. While recent CL work has advanced in mitigating catastrophic forgetting and improving knowledge transfer, the practical requirements of energy and memory efficiency for tabular data streams remain underexplored. In particular, existing CL solutions mostly depend on replay mechanisms whose buffers grow over time and exacerbate resource costs. We propose a context-aware incremental Multi-Layer Perceptron (IMLP), a compact continual learner for tabular data streams. IMLP incorporates a windowed scaled dot-product attention over a sliding latent feature buffer, enabling constant-size memory and avoiding storing raw data. The attended context is concatenated with current features and processed by shared feed-forward layers, yielding lightweight per-segment updates. To assess practical deployability, we introduce NetScore-T, a tunable metric coupling balanced accuracy with energy for Pareto-aware comparison across models and datasets. IMLP achieves up to 27.6\times higher energy efficiency than TabNet and 85.5\times higher than TabPFN, while maintaining competitive average accuracy. Overall, IMLP provides an easy-to-deploy, energy-efficient alternative to full retraining for tabular data streams. Subjects: Machine Learning (cs.LG) Cite as: arXiv:2510.04660 [cs.LG] (or arXiv:2510.04660v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2510.04660 Focus to learn more arXiv-issued DOI via DataCite (pending registration)

[LG-30] Compressed Concatenation of Small Embedding Models

链接: https://arxiv.org/abs/2510.04626
作者: Mohamed Ayoub Ben Ayad,Michael Dinzinger,Kanishka Ghosh Dastidar,Jelena Mitrovic,Michael Granitzer
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-31] Forecasting-Based Biomedical Time-series Data Synthesis for Open Data and Robust AI

链接: https://arxiv.org/abs/2510.04622
作者: Youngjoon Lee,Seongmin Cho,Yehhyun Jo,Jinu Gong,Hyunjoo Jenny Lee,Joonhyuk Kang
类目: Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: Under Review

点击查看摘要

[LG-32] Closed-Form Last Layer Optimization

链接: https://arxiv.org/abs/2510.04606
作者: Alexandre Galashov,Nathaël Da Costa,Liyuan Xu,Philipp Hennig,Arthur Gretton
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-33] Data-Driven Adaptive PID Control Based on Physics-Informed Neural Networks

链接: https://arxiv.org/abs/2510.04591
作者: Junsei Ito,Yasuaki Wasa
类目: ystems and Control (eess.SY); Machine Learning (cs.LG)
*备注: This work has been submitted to the IEEE Transactions on Control Systems Technology for possible publication

点击查看摘要

[LG-34] Improved probabilistic regression using diffusion models

链接: https://arxiv.org/abs/2510.04583
作者: Carlo Kneissl,Christopher Bülte,Philipp Scholl,Gitta Kutyniok
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-35] Busemann Functions in the Wasserstein Space: Existence Closed-Forms and Applications to Slicing

链接: https://arxiv.org/abs/2510.04579
作者: Clément Bonet,Elsa Cazelles,Lucas Drumetz,Nicolas Courty
类目: Machine Learning (cs.LG); Metric Geometry (math.MG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-36] Language Model Based Text-to-Audio Generation: Anti-Causally Aligned Collaborative Residual Transformers EMNLP2025

链接: https://arxiv.org/abs/2510.04577
作者: Juncheng Wang,Chao Xu,Cheng Yu,Zhe Hu,Haoyu Xie,Guoqi Yu,Lei Shang,Shujun Wang
类目: ound (cs.SD); Machine Learning (cs.LG); Multimedia (cs.MM); Audio and Speech Processing (eess.AS)
*备注: Accepted to EMNLP 2025

点击查看摘要

[LG-37] Stochastic Approximation Methods for Distortion Risk Measure Optimization

链接: https://arxiv.org/abs/2510.04563
作者: Jinyang Jiang,Bernd Heidergott,Jiaqiao Hu,Yijie Peng
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:

点击查看摘要

[LG-38] Challenger-Based Combinatorial Bandits for Subcarrier Selection in OFDM Systems

链接: https://arxiv.org/abs/2510.04559
作者: Mohsen Amiri,V Venktesh,Sindri Magnússon
类目: Machine Learning (cs.LG)
*备注: 6 pages

点击查看摘要

[LG-39] ail-Safe Hedging: Explainable Risk-Sensitive Reinforcement Learning with a White-Box CBF–QP Safety Layer in Arbitrag e-Free Markets

链接: https://arxiv.org/abs/2510.04555
作者: Jian’an Zhang
类目: Machine Learning (cs.LG); Trading and Market Microstructure (q-fin.TR)
*备注: 32 pages including appendices; 5 figures. Primary subject class: q-fin.TR. Cross-lists: cs.LG; q-fin.RM

点击查看摘要

[LG-40] Graph-based Tabular Deep Learning Should Learn Feature Interactions Not Just Make Predictions NEURIPS2025

链接: https://arxiv.org/abs/2510.04543
作者: Elias Dubbeldam,Reza Mohammadi,Marit Schoonhoven,S. Ilker Birbil
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 9 pages, 6 figures, submitted to position track NeurIPS 2025

点击查看摘要

[LG-41] Demystifying MaskGIT Sampler and Beyond: Adaptive Order Selection in Masked Diffusion

链接: https://arxiv.org/abs/2510.04525
作者: Satoshi Hayakawa,Yuhta Takida,Masaaki Imaizumi,Hiromi Wakaki,Yuki Mitsufuji
类目: Machine Learning (cs.LG); Probability (math.PR); Machine Learning (stat.ML)
*备注: 23 pages

点击查看摘要

Abstract:Masked diffusion models have shown promising performance in generating high-quality samples in a wide range of domains, but accelerating their sampling process remains relatively underexplored. To investigate efficient samplers for masked diffusion, this paper theoretically analyzes the MaskGIT sampler for image modeling, revealing its implicit temperature sampling mechanism. Through this analysis, we introduce the “moment sampler,” an asymptotically equivalent but more tractable and interpretable alternative to MaskGIT, which employs a “choose-then-sample” approach by selecting unmasking positions before sampling tokens. In addition, we improve the efficiency of choose-then-sample algorithms through two key innovations: a partial caching technique for transformers that approximates longer sampling trajectories without proportional computational cost, and a hybrid approach formalizing the exploration-exploitation trade-off in adaptive unmasking. Experiments in image and text domains demonstrate our theory as well as the efficiency of our proposed methods, advancing both theoretical understanding and practical implementation of masked diffusion samplers.

[LG-42] Wavelet Predictive Representations for Non-Stationary Reinforcement Learning

链接: https://arxiv.org/abs/2510.04507
作者: Min Wang,Xin Li,Ye He,Yao-Hui Li,Hasnaa Bennis,Riashat Islam,Mingzhong Wang
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-43] Causality-aware Graph Aggregation Weight Estimator for Popularity Debiasing in Top-K Recommendation CIKM2025

链接: https://arxiv.org/abs/2510.04502
作者: Yue Que,Yingyi Zhang,Xiangyu Zhao,Chen Ma
类目: Information Retrieval (cs.IR); Machine Learning (cs.LG)
*备注: Accepted by CIKM 2025

点击查看摘要

[LG-44] Expand Neurons Not Parameters

链接: https://arxiv.org/abs/2510.04500
作者: Linghao Kong,Inimai Subramanian,Yonadav Shavit,Micah Adler,Dan Alistarh,Nir Shavit
类目: Machine Learning (cs.LG)
*备注: 10 pages, 6 figures

点击查看摘要

[LG-45] Deep vs. Shallow: Benchmarking Physics-Informed Neural Architectures on the Biharmonic Equation NEURIPS

链接: https://arxiv.org/abs/2510.04490
作者: Akshay Govind Srinivasan,Vikas Dwivedi,Balaji Srinivasan
类目: Computational Engineering, Finance, and Science (cs.CE); Emerging Technologies (cs.ET); Machine Learning (cs.LG)
*备注: 16 Pages, 7 Figures and 1 Table. Submitted and accepted at Machine Learning and the Physical Sciences Workshop at the 39th conference on Neural Information Processing Systems (NeurIPS)

点击查看摘要

[LG-46] Forking-Sequences

链接: https://arxiv.org/abs/2510.04487
作者: Willa Potosnak,Malcolm Wolff,Boris Oreshkin,Mengfei Cao,Michael W. Mahoney,Dmitry Efimov,Kin G. Olivares
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-47] Domain Generalization: A Tale of Two ERMs

链接: https://arxiv.org/abs/2510.04441
作者: Yilun Zhu,Naihao Deng,Naichen Shi,Aditya Gangrade,Clayton Scott
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-48] Fractional Heat Kernel for Semi-Supervised Graph Learning with Small Training Sample Size

链接: https://arxiv.org/abs/2510.04440
作者: Farid Bozorgnia,Vyacheslav Kungurtsev,Shirali Kadyrov,Mohsen Yousefnezhad
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-49] rade-off in Estimating the Number of Byzantine Clients in Federated Learning

链接: https://arxiv.org/abs/2510.04432
作者: Ziyi Chen,Su Zhang,Heng Huang
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:

点击查看摘要

[LG-50] Achieve Performatively Optimal Policy for Performative Reinforcement Learning

链接: https://arxiv.org/abs/2510.04430
作者: Ziyi Chen,Heng Huang
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:

点击查看摘要

Abstract:Performative reinforcement learning is an emerging dynamical decision making framework, which extends reinforcement learning to the common applications where the agent’s policy can change the environmental dynamics. Existing works on performative reinforcement learning only aim at a performatively stable (PS) policy that maximizes an approximate value function. However, there is a provably positive constant gap between the PS policy and the desired performatively optimal (PO) policy that maximizes the original value function. In contrast, this work proposes a zeroth-order Frank-Wolfe algorithm (0-FW) algorithm with a zeroth-order approximation of the performative policy gradient in the Frank-Wolfe framework, and obtains \textbfthe first polynomial-time convergence to the desired PO policy under the standard regularizer dominance condition. For the convergence analysis, we prove two important properties of the nonconvex value function. First, when the policy regularizer dominates the environmental shift, the value function satisfies a certain gradient dominance property, so that any stationary point (not PS) of the value function is a desired PO. Second, though the value function has unbounded gradient, we prove that all the sufficiently stationary points lie in a convex and compact policy subspace \Pi_\Delta , where the policy value has a constant lower bound \Delta0 and thus the gradient becomes bounded and Lipschitz continuous. Experimental results also demonstrate that our 0-FW algorithm is more effective than the existing algorithms in finding the desired PO policy.

[LG-51] Scale-Invariant Regret Matching and Online Learning with Optimal Convergence: Bridging Theory and Practice in Zero-Sum Games

链接: https://arxiv.org/abs/2510.04407
作者: Brian Hu Zhang,Ioannis Anagnostides,Tuomas Sandholm
类目: Computer Science and Game Theory (cs.GT); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-52] SSM-CGM: Interpretable State-Space Forecasting Model of Continuous Glucose Monitoring for Personalized Diabetes Management

链接: https://arxiv.org/abs/2510.04386
作者: Shakson Isaac,Yentl Collin,Chirag Patel
类目: Machine Learning (cs.LG)
*备注: Shakson Isaac and Yentl Collin contributed equally

点击查看摘要

[LG-53] Score-based Greedy Search for Structure Identification of Partially Observed Linear Causal Models

链接: https://arxiv.org/abs/2510.04378
作者: Xinshuai Dong,Ignavier Ng,Haoyue Dai,Jiaqi Sun,Xiangchen Song,Peter Spirtes,Kun Zhang
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-54] Categorical Invariants of Learning Dynamics

链接: https://arxiv.org/abs/2510.04376
作者: Abdulrahman Tamim
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-55] Quantifying Ambiguity in Categorical Annotations: A Measure and Statistical Inference Framework

链接: https://arxiv.org/abs/2510.04366
作者: Christopher Klugmann,Daniel Kondermann
类目: Machine Learning (cs.LG)
*备注: Preprint, 20 pages in total, 7 figures

点击查看摘要

[LG-56] From News to Returns: A Granger-Causal Hypergraph Transformer on the Sphere

链接: https://arxiv.org/abs/2510.04357
作者: Anoushka Harit,Zhongtian Sun,Jongmin Yu
类目: Machine Learning (cs.LG); Computational Finance (q-fin.CP)
*备注: 6th ACM International Conference on AI in Finance

点击查看摘要

[LG-57] Environment-Aware Indoor LoRaWAN Path Loss: Parametric Regression Comparisons Shadow Fading and Calibrated Fade Margins

链接: https://arxiv.org/abs/2510.04346
作者: Nahshon Mokua Obiri,Kristof Van Laerhoven
类目: Networking and Internet Architecture (cs.NI); Machine Learning (cs.LG); Signal Processing (eess.SP); Numerical Analysis (math.NA)
*备注: Code: this https URL

点击查看摘要

[LG-58] Learning to Predict Chaos: Curriculum-Driven Training for Robust Forecasting of Chaotic Dynamics

链接: https://arxiv.org/abs/2510.04342
作者: Harshil Vejendla
类目: Machine Learning (cs.LG)
*备注: MIT URTC Technical Paper (Oral), 5 pages, 4 figures

点击查看摘要

[LG-59] Arithmetic-Mean μP for Modern Architectures: A Unified Learning-Rate Scale for CNNs and ResNets ICLR2026

链接: https://arxiv.org/abs/2510.04327
作者: Haosong Zhang,Shenxi Wu,Yichi Zhang,Wei Lin
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: Preprint. Under review at ICLR 2026

点击查看摘要

[LG-60] FoilDiff: A Hybrid Transformer Backbone for Diffusion-based Modelling of 2D Airfoil Flow Fields

链接: https://arxiv.org/abs/2510.04325
作者: Kenechukwu Ogbuagu,Sepehr Maleki,Giuseppe Bruni,Senthil Krishnababu
类目: Machine Learning (cs.LG); Fluid Dynamics (physics.flu-dyn)
*备注:

点击查看摘要

[LG-61] owards Fast Option Pricing PDE Solvers Powered by PIELM

链接: https://arxiv.org/abs/2510.04322
作者: Akshay Govind Srinivasan,Anuj Jagannath Said,Sathwik Pentela,Vikas Dwivedi,Balaji Srinivasan
类目: Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注: 6 Pages, 5 Figures, 3 Tables

点击查看摘要

[LG-62] Crash Severity Prediction Using Deep Learning Approaches: A Hybrid CNN-RNN Framework

链接: https://arxiv.org/abs/2510.04316
作者: Sahar Koohfar
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-63] Activation Steering with a Feedback Controller

链接: https://arxiv.org/abs/2510.04309
作者: Dung V. Nguyen,Hieu M. Vu,Nhi Y. Pham,Lei Zhang,Tan M. Nguyen
类目: Machine Learning (cs.LG)
*备注: 9 pages in the main text. Under Review

点击查看摘要

[LG-64] HoRA: Cross-Head Low-Rank Adaptation with Joint Hypernetworks

链接: https://arxiv.org/abs/2510.04295
作者: Nghiem T. Diep,Dung Le,Tuan Truong,Tan Dinh,Huy Nguyen,Nhat Ho
类目: Machine Learning (cs.LG)
*备注: Nghiem T. Diep, Dung Le, and Tuan Truong contributed equally to this work

点击查看摘要

Abstract:Low-Rank Adaptation (LoRA) is a parameter-efficient fine-tuning (PEFT) technique that adapts large pre-trained models by adding low-rank matrices to their weight updates. However, in the context of fine-tuning multi-head self-attention (MHA), LoRA has been employed to adapt each attention head separately, thereby overlooking potential synergies across different heads. To mitigate this issue, we propose a novel Hyper-shared Low-Rank Adaptation (HoRA) method, which utilizes joint hypernetworks to generate low-rank matrices across attention heads. By coupling their adaptation through a shared generator, HoRA encourages cross-head information sharing, and thus directly addresses the aforementioned limitation of LoRA. By comparing LoRA and HoRA through the lens of hierarchical mixture of experts, our theoretical findings reveal that the latter achieves superior sample efficiency to the former. Furthermore, through extensive experiments across diverse language and vision benchmarks, we demonstrate that HoRA outperforms LoRA and other PEFT methods while requiring only a marginal increase in the number of trainable parameters.

[LG-65] Influence branching for learning to solve mixed-integer programs online

链接: https://arxiv.org/abs/2510.04273
作者: Paul Strang,Zacharie Alès,Côme Bissuel,Olivier Juan,Safia Kedad-Sidhoum,Emmanuel Rachelson
类目: Machine Learning (cs.LG)
*备注: 11 pages

点击查看摘要

[LG-66] runcated Kernel Stochastic Gradient Descent with General Losses and Spherical Radial Basis Functions

链接: https://arxiv.org/abs/2510.04237
作者: Jinhui Bai,Andreas Christmann,Lei Shi
类目: Machine Learning (cs.LG)
*备注: 54 pages, 20 figures

点击查看摘要

[LG-67] Adaptive Federated Learning via Dynamical System Model

链接: https://arxiv.org/abs/2510.04203
作者: Aayushya Agarwal,Larry Pileggi,Gauri Joshi
类目: Machine Learning (cs.LG); Systems and Control (eess.SY)
*备注:

点击查看摘要

[LG-68] Spectral Alignment as Predictor of Loss Explosion in Neural Network Training

链接: https://arxiv.org/abs/2510.04202
作者: Haiquan Qiu,You Wu,Yingjie Tan,Yaqing Wang,Quanming Yao
类目: Machine Learning (cs.LG)
*备注: 18 pages, 8 figures

点击查看摘要

[LG-69] ObCLIP: Oblivious CLoud-Device Hybrid Image Generation with Privacy Preservation NEURIPS2025

链接: https://arxiv.org/abs/2510.04153
作者: Haoqi Wu,Wei Dai,Ming Xu,Li Wang,Qiang Yan
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: Accepted by NeurIPS 2025

点击查看摘要

Abstract:Diffusion Models have gained significant popularity due to their remarkable capabilities in image generation, albeit at the cost of intensive computation requirement. Meanwhile, despite their widespread deployment in inference services such as Midjourney, concerns about the potential leakage of sensitive information in uploaded user prompts have arisen. Existing solutions either lack rigorous privacy guarantees or fail to strike an effective balance between utility and efficiency. To bridge this gap, we propose ObCLIP, a plug-and-play safeguard that enables oblivious cloud-device hybrid generation. By oblivious, each input prompt is transformed into a set of semantically similar candidate prompts that differ only in sensitive attributes (e.g., gender, ethnicity). The cloud server processes all candidate prompts without knowing which one is the real one, thus preventing any prompt leakage. To mitigate server cost, only a small portion of denoising steps is performed upon the large cloud model. The intermediate latents are then sent back to the client, which selects the targeted latent and completes the remaining denoising using a small device model. Additionally, we analyze and incorporate several cache-based accelerations that leverage temporal and batch redundancy, effectively reducing computation cost with minimal utility degradation. Extensive experiments across multiple datasets demonstrate that ObCLIP provides rigorous privacy and comparable utility to cloud models with slightly increased server cost.

[LG-70] Efficient Manifold-Constrained Neural ODE for High-Dimensional Datasets IJCNN

链接: https://arxiv.org/abs/2510.04138
作者: Muhao Guo,Haoran Li,Yang Weng
类目: Machine Learning (cs.LG)
*备注: 8 pages; 7 figures; conference IJCNN

点击查看摘要

[LG-71] Modeling Time Series Dynamics with Fourier Ordinary Differential Equations

链接: https://arxiv.org/abs/2510.04133
作者: Muhao Guo,Yang Weng
类目: Machine Learning (cs.LG)
*备注: 8 pages, 7 figures, conference

点击查看摘要

[LG-72] On the Statistical Query Complexity of Learning Semiautomata: a Random Walk Approach

链接: https://arxiv.org/abs/2510.04115
作者: George Giapitzakis,Kimon Fountoulakis,Eshaan Nichani,Jason D. Lee
类目: Machine Learning (cs.LG)
*备注: 42 pages

点击查看摘要

[LG-73] Wasserstein projection distance for fairness testing of regression models

链接: https://arxiv.org/abs/2510.04114
作者: Wanxin Li,Yongjin P. Park,Khanh Dao Duc
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-74] Can Linear Probes Measure LLM Uncertainty?

链接: https://arxiv.org/abs/2510.04108
作者: Ramzi Dakhmouche,Adrien Letellier,Hossein Gorji
类目: Machine Learning (cs.LG); Numerical Analysis (math.NA); Statistics Theory (math.ST)
*备注:

点击查看摘要

[LG-75] Why Cannot Neural Networks Master Extrapolation? Insights from Physical Laws

链接: https://arxiv.org/abs/2510.04102
作者: Ramzi Dakhmouche,Hossein Gorji
类目: Machine Learning (cs.LG); Numerical Analysis (math.NA); Probability (math.PR)
*备注:

点击查看摘要

[LG-76] RLRF: Competitive Search Agent Design via Reinforcement Learning from Ranker Feedback

链接: https://arxiv.org/abs/2510.04096
作者: Tommy Mordo,Sagie Dekel,Omer Madmon,Moshe Tennenholtz,Oren Kurland
类目: Information Retrieval (cs.IR); Computer Science and Game Theory (cs.GT); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-77] Rethinking Consistent Multi-Label Classification under Inexact Supervision

链接: https://arxiv.org/abs/2510.04091
作者: Wei Wang,Tianhao Ma,Ming-Kun Xie,Gang Niu,Masashi Sugiyama
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-78] Sharp Lower Bounds for Linearized ReLUk Approximation on the Sphere

链接: https://arxiv.org/abs/2510.04060
作者: Tong Mao,Jinchao Xu
类目: Numerical Analysis (math.NA); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-79] Variational Diffusion Unlearning: A Variational Inference Framework for Unlearning in Diffusion Models under Data Constraints

链接: https://arxiv.org/abs/2510.04058
作者: Subhodip Panda,MS Varun,Shreyans Jain,Sarthak Kumar Maharana,Prathosh A.P
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-80] Adaptive kernel-density approach for imbalanced binary classification

链接: https://arxiv.org/abs/2510.04046
作者: Kotaro J. Nishimura,Yuichi Sakumura,Kazushi Ikeda
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-81] Multi-Class Support Vector Machine with Differential Privacy NEURIPS2025

链接: https://arxiv.org/abs/2510.04027
作者: Jinseong Park,Yujin Choi,Jaewook Lee
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注: NeurIPS 2025

点击查看摘要

[LG-82] Incorporating Multivariate Consistency in ML-Based Weather Forecasting with Latent-space Constraints

链接: https://arxiv.org/abs/2510.04006
作者: Hang Fan,Yi Xiao,Yongquan Qu,Fenghua Ling,Ben Fei,Lei Bai,Pierre Gentine
类目: Machine Learning (cs.LG); Chaotic Dynamics (nlin.CD); Atmospheric and Oceanic Physics (physics.ao-ph)
*备注:

点击查看摘要

[LG-83] Multi-Modal Multi-Task Semantic Communication: A Distributed Information Bottleneck Perspective

链接: https://arxiv.org/abs/2510.04000
作者: Yujie Zhou,Yiwei Liao,Cheng Peng,Yong Xiao,Yingyu Li
类目: Information Theory (cs.IT); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-84] ICEPool: Enhancing Graph Pooling Networks with Inter-cluster Connectivity

链接: https://arxiv.org/abs/2510.03987
作者: Michael Yang
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-85] Beyond Static Evaluation: Rethinking the Assessment of Personalized Agent Adaptability in Information Retrieval

链接: https://arxiv.org/abs/2510.03984
作者: Kirandeep Kaur,Preetam Prabhu Srikar Dammu,Hideo Joho,Chirag Shah
类目: Information Retrieval (cs.IR); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-86] Beyond Softmax: A New Perspective on Gradient Bandits

链接: https://arxiv.org/abs/2510.03979
作者: Emerson Melo,David Müller
类目: Machine Learning (cs.LG); Theoretical Economics (econ.TH)
*备注:

点击查看摘要

[LG-87] Early-Warning of Thunderstorm-Driven Power Outages with a Two-Stage Machine Learning Model

链接: https://arxiv.org/abs/2510.03959
作者: Iryna Stanishevska
类目: Machine Learning (cs.LG)
*备注: 23 pages (main), 70 pages incl. appendices; figures tables as in manuscript. Code (main figure, synthetic data): this https URL License: CC BY 4.0 (preprint)

点击查看摘要

[LG-88] Optimizing Resources for On-the-Fly Label Estimation with Multiple Unknown Medical Experts

链接: https://arxiv.org/abs/2510.03954
作者: Tim Bary,Tiffanie Godelaine,Axel Abels,Benoît Macq
类目: Machine Learning (cs.LG)
*备注: 7 pages, 3 figures, 3 tables, Accepted at IEEE BHI 2025

点击查看摘要

[LG-89] What Is The Performance Ceiling of My Classifier? Utilizing Category-Wise Influence Functions for Pareto Frontier Analysis

链接: https://arxiv.org/abs/2510.03950
作者: Shahriar Kabir Nahin,Wenxiao Xiao,Joshua Liu,Anshuman Chhabra,Hongfu Liu
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-90] On the Empirical Power of Goodness-of-Fit Tests in Watermark Detection NEURIPS2025

链接: https://arxiv.org/abs/2510.03944
作者: Weiqing He,Xiang Li,Tianqi Shang,Li Shen,Weijie Su,Qi Long
类目: Machine Learning (cs.LG)
*备注: Accepted at NeurIPS 2025 as a spotlight

点击查看摘要

Abstract:Large language models (LLMs) raise concerns about content authenticity and integrity because they can generate human-like text at scale. Text watermarks, which embed detectable statistical signals into generated text, offer a provable way to verify content origin. Many detection methods rely on pivotal statistics that are i.i.d. under human-written text, making goodness-of-fit (GoF) tests a natural tool for watermark detection. However, GoF tests remain largely underexplored in this setting. In this paper, we systematically evaluate eight GoF tests across three popular watermarking schemes, using three open-source LLMs, two datasets, various generation temperatures, and multiple post-editing methods. We find that general GoF tests can improve both the detection power and robustness of watermark detectors. Notably, we observe that text repetition, common in low-temperature settings, gives GoF tests a unique advantage not exploited by existing methods. Our results highlight that classic GoF tests are a simple yet powerful and underused tool for watermark detection in LLMs.

[LG-91] ransductive and Learning-Augmented Online Regression

链接: https://arxiv.org/abs/2510.03917
作者: Vinod Raman,Shenghao Xie,Samson Zhou
类目: Machine Learning (cs.LG); Data Structures and Algorithms (cs.DS)
*备注:

点击查看摘要

[LG-92] Generalized Fitted Q-Iteration with Clustered Data

链接: https://arxiv.org/abs/2510.03912
作者: Liyuan Hu,Jitao Wang,Zhenke Wu,Chengchun Shi
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-93] HEMIS: Unlocking Pretrained Knowledge with Foundation Model Embeddings for Anomaly Detection in Time Series IJCAI’25

链接: https://arxiv.org/abs/2510.03911
作者: Yadav Mahesh Lorik,Kaushik Sarveswaran,Nagaraj Sundaramahalingam,Aravindakumar Venugopalan
类目: Machine Learning (cs.LG)
*备注: Oral Presentation. AI4TS Workshop, IJCAI’25

点击查看摘要

[LG-94] LLM as an Algorithmist: Enhancing Anomaly Detectors via Programmatic Synthesis

链接: https://arxiv.org/abs/2510.03904
作者: Hangting Ye,Jinmeng Li,He Zhao,Mingchen Zhuge,Dandan Guo,Yi Chang,Hongyuan Zha
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-95] Fair Minimum Labeling: Efficient Temporal Network Activations for Reachability and Equity NEURIPS2025

链接: https://arxiv.org/abs/2510.03899
作者: Lutz Oettershagen,Othon Michail
类目: ocial and Information Networks (cs.SI); Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG)
*备注: Accepted at NeurIPS 2025

点击查看摘要

[LG-96] BONSAI: Structure-exploiting robust Bayesian optimization for networked black-box systems under uncertainty

链接: https://arxiv.org/abs/2510.03893
作者: Akshay Kudva,Joel A. Paulson
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注: Published in Computers and Chemical Engineering, 2025

点击查看摘要

[LG-97] On Provable Benefits of Muon in Federated Learning

链接: https://arxiv.org/abs/2510.03866
作者: Xinwen Zhang,Hongchang Gao
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-98] On Using Large Language Models to Enhance Clinically-Driven Missing Data Recovery Algorithms in Electronic Health Records

链接: https://arxiv.org/abs/2510.03844
作者: Sarah C. Lotspeich,Abbey Collins,Brian J. Wells,Ashish K. Khanna,Joseph Rigdon,Lucy D’Agostino McGowan
类目: Machine Learning (cs.LG); Applications (stat.AP); Methodology (stat.ME)
*备注:

点击查看摘要

[LG-99] Smart Paste: Automatically Fixing Copy/Paste for Google Developers

链接: https://arxiv.org/abs/2510.03843
作者: Vincent Nguyen,Guilherme Herzog,José Cambronero,Marcus Revaj,Aditya Kini,Alexander Frömmgen,Maxim Tabachnyk
类目: oftware Engineering (cs.SE); Human-Computer Interaction (cs.HC); Machine Learning (cs.LG)
*备注: 11 pages

点击查看摘要

[LG-100] chnical note on Sequential Test-Time Adaptation via Martingale-Driven Fisher Prompting

链接: https://arxiv.org/abs/2510.03839
作者: Behraj Khan,Tahir Qasim Syed
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-101] chnical note on Fisher Information for Robust Federated Cross-Validation

链接: https://arxiv.org/abs/2510.03838
作者: Behraj Khan,Tahir Qasim Syed
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-102] Pilot Contamination Attacks Detection with Machine Learning for Multi-User Massive MIMO

链接: https://arxiv.org/abs/2510.03831
作者: Pedro Ivo da Cruz,Dimitri Silva,Tito Spadini,Ricardo Suyama,Murilo Bellezoni Loiola
类目: Cryptography and Security (cs.CR); Information Theory (cs.IT); Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: This version of the article has been accepted for publication, after peer review and is subject to Springer Nature’s AM terms of use, but is not the Version of Record and does not reflect post-acceptance improvements, or any corrections. The Version of Record is available online at: this https URL

点击查看摘要

[LG-103] HOFLON: Hybrid Offline Learning and Online Optimization for Process Start-Up and Grade-Transition Control

链接: https://arxiv.org/abs/2510.03830
作者: Alex Durkin,Jasper Stolte,Mehmet Mercangöz
类目: Machine Learning (cs.LG); Systems and Control (eess.SY); Machine Learning (stat.ML)
*备注: 31 pages, 15 figures, submitted to Computers and Chemical Engineering

点击查看摘要

[LG-104] Distributed Area Coverag e with High Altitude Balloons Using Multi-Agent Reinforcement Learning

链接: https://arxiv.org/abs/2510.03823
作者: Adam Haroon,Tristan Schuler
类目: Machine Learning (cs.LG); Multiagent Systems (cs.MA); Robotics (cs.RO)
*备注:

点击查看摘要

[LG-105] ROLL: Trust Regions improve Reinforcement Learning for Large Language Models

链接: https://arxiv.org/abs/2510.03817
作者: Philipp Becker,Niklas Freymuth,Serge Thilges,Fabian Otto,Gerhard Neumann
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-106] A Trustworthy Industrial Fault Diagnosis Architecture Integrating Probabilistic Models and Large Language Models

链接: https://arxiv.org/abs/2510.03815
作者: Yue wu
类目: ystems and Control (eess.SY); Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: 1tables,6 figs,11pages

点击查看摘要

[LG-107] Curriculum-Augmented GFlowNets For mRNA Sequence Generation

链接: https://arxiv.org/abs/2510.03811
作者: Aya Laajil,Abduragim Shtanchaev,Sajan Muhammad,Eric Moulines,Salem Lahlou
类目: Machine Learning (cs.LG); Quantitative Methods (q-bio.QM)
*备注:

点击查看摘要

[LG-108] Cellular Learning: Scattered Data Regression in High Dimensions via Voronoi Cells

链接: https://arxiv.org/abs/2510.03810
作者: Shankar Prasad Sastry
类目: Computational Geometry (cs.CG); Machine Learning (cs.LG)
*备注: 15 pages + 2 pages references; 3 figures; 4 tables; 1 algorithm

点击查看摘要

[LG-109] Robust Batched Bandits

链接: https://arxiv.org/abs/2510.03798
作者: Yunwen Guo,Yunlun Shu,Gongyi Zhuo,Tianyu Wang
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 39 pages

点击查看摘要

[LG-110] Allocation of Parameters in Transformers

链接: https://arxiv.org/abs/2510.03784
作者: Ruoxi Yu,Haotian Jiang,Jingpu Cheng,Penghao Yu,Qianxiao Li,Zhong Li
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-111] Merge and Guide: Unifying Model Merging and Guided Decoding for Controllable Multi-Objective Generation

链接: https://arxiv.org/abs/2510.03782
作者: Guofu Xie,Chen Zhang,Xiao Zhang,Yunsheng Shi,Ting Yao,Jun Xu
类目: Machine Learning (cs.LG)
*备注: Work in progress

点击查看摘要

[LG-112] rajectory prediction for heterogeneous agents : A performance analysis on small and imbalanced datasets

链接: https://arxiv.org/abs/2510.03776
作者: Tiago Rodrigues de Almeida,Yufei Zhu,Andrey Rudenko,Tomasz P. Kucner,Johannes A. Stork,Martin Magnusson,Achim J. Lilienthal
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: This paper has been accepted to the IEEE Robotics and Automation Letters journal and presented at the 40th Anniversary of the IEEE International Conference on Robotics and Automation, which was held in Rotterdam, Netherlands on 23-26 September, 2024

点击查看摘要

[LG-113] Neural Low-Discrepancy Sequences

链接: https://arxiv.org/abs/2510.03745
作者: Michael Etienne Van Huffel,Nathan Kirk,Makram Chahine,Daniela Rus,T. Konstantin Rusch
类目: Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注:

点击查看摘要

[LG-114] Lightweight and Generalizable Acoustic Scene Representations via Contrastive Fine-Tuning and Distillation

链接: https://arxiv.org/abs/2510.03728
作者: Kuang Yuan,Yang Gao,Xilin Li,Xinhao Mei,Syavosh Zadissa,Tarun Pruthi,Saeed Bagheri Sereshki
类目: ound (cs.SD); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS); Signal Processing (eess.SP)
*备注:

点击查看摘要

[LG-115] Personalized federated prototype learning in mixed heterogeneous data scenarios

链接: https://arxiv.org/abs/2510.03726
作者: Jiahao Zeng,Wolong Xing,Liangtao Shi,Xin Huang,Jialin Wang,Zhile Cao,Zhenkui Shi
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-116] Balancing Interpretability and Performance in Reinforcement Learning: An Adaptive Spectral Based Linear Approach

链接: https://arxiv.org/abs/2510.03722
作者: Qianxin Yi,Shao-Bo Lin,Jun Fan,Yao Wang
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-117] From Moments to Models: Graphon Mixture-Aware Mixup and Contrastive Learning

链接: https://arxiv.org/abs/2510.03690
作者: Ali Azizpour,Reza Ramezanpour,Ashutosh Sabharwal,Santiago Segarra
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-118] Group Policy Gradient

链接: https://arxiv.org/abs/2510.03679
作者: Junhua Chen,Zixi Zhang,Hantao Zhong,Rika Antonova
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-119] owards Sampling Data Structures for Tensor Products in Turnstile Streams

链接: https://arxiv.org/abs/2510.03678
作者: Zhao Song,Shenghao Xie,Samson Zhou
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-120] Optimising Battery Energy Storag e System Trading via Energy Market Operator Price Forecast

链接: https://arxiv.org/abs/2510.03657
作者: Aymeric Fabre
类目: Machine Learning (cs.LG); Systems and Control (eess.SY)
*备注:

点击查看摘要

[LG-121] SAFA-SNN: Sparsity-Aware On-Device Few-Shot Class-Incremental Learning with Fast-Adaptive Structure of Spiking Neural Network

链接: https://arxiv.org/abs/2510.03648
作者: Huijing Zhang,Muyang Cao,Linshan Jiang,Xin Du,Di Yu,Changze Lv,Shuiguang Deng
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-122] In-Vivo Training for Deep Brain Stimulation

链接: https://arxiv.org/abs/2510.03643
作者: Nicholas Carter,Arkaprava Gupta,Prateek Ganguli,Benedikt Dietrich,Vibhor Krishna,Samarjit Chakraborty
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-123] Explore the Loss space with Hill-ADAM

链接: https://arxiv.org/abs/2510.03613
作者: Meenakshi Manikandan,Leilani Gilpin
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 14-15 pages

点击查看摘要

[LG-124] MECKD: Deep Learning-Based Fall Detection in Multilayer Mobile Edge Computing With Knowledge Distillation

链接: https://arxiv.org/abs/2510.03601
作者: Wei-Lung Mao,Chun-Chi Wang,Po-Heng Chou,Kai-Chun Liu,Yu Tsao
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC); Networking and Internet Architecture (cs.NI); Signal Processing (eess.SP)
*备注: 15 pages, 7 figures, and published in IEEE Sensors Journal

点击查看摘要

[LG-125] FieldFormer: Physics-Informed Transformers for Spatio-Temporal Field Reconstruction from Sparse Sensors

链接: https://arxiv.org/abs/2510.03589
作者: Ankit Bhardwaj,Ananth Balashankar,Lakshminarayanan Subramanian
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-126] BEKAN: Boundary condition-guaranteed evolutionary Kolmogorov-Arnold networks with radial basis functions for solving PDE problems

链接: https://arxiv.org/abs/2510.03576
作者: Bongseok Kim,Jiahao Zhang,Guang Lin
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 29 pages, 22 figures

点击查看摘要

[LG-127] CrossLag: Predicting Major Dengue Outbreaks with a Domain Knowledge Informed Transformer

链接: https://arxiv.org/abs/2510.03566
作者: Ashwin Prabu,Nhat Thanh Tran,Guofa Zhou,Jack Xin
类目: Machine Learning (cs.LG); Computers and Society (cs.CY)
*备注: © 2025 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works

点击查看摘要

[LG-128] Sequential decoder training for improved latent space dynamics identification

链接: https://arxiv.org/abs/2510.03535
作者: William Anderson,Seung Whan Chung,Youngsoo Choi
类目: Machine Learning (cs.LG); Numerical Analysis (math.NA); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-129] Long-Term Mapping of the Douro River Plume with Multi-Agent Reinforcement Learning

链接: https://arxiv.org/abs/2510.03534
作者: Nicolò Dal Fabbro,Milad Mesbahi,Renato Mendes,João Borges de Sousa,George J. Pappas
类目: Multiagent Systems (cs.MA); Machine Learning (cs.LG); Systems and Control (eess.SY); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-130] RAPID: An Efficient Reinforcement Learning Algorithm for Small Language Models

链接: https://arxiv.org/abs/2510.03515
作者: Lianghuan Huang,Sagnik Anupam,Insup Lee,Shuo Li,Osbert Bastani
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

Abstract:Reinforcement learning (RL) has emerged as a promising strategy for finetuning small language models (SLMs) to solve targeted tasks such as math and coding. However, RL algorithms tend to be resource-intensive, taking a significant amount of time to train. We propose RAPID, a novel RL algorithm that can substantially reduce the running time of RL. Our key insight is that RL tends to be costly due to the need to perform both inference and backpropagation during training. To maximize use of computational resources, our algorithm performs inference in large batches, and then performs off-policy policy gradient updates in mini-batches. For off-policy updates, we incorporate group advantage estimation into the policy gradient algorithm, and derive an importance weighted estimator to correct for the bias arising from off-policy learning. Our experiments demonstrate that our algorithm can reduce running time by 11%-34% on three benchmarks compared to state-of-the-art RL algorithms while maintaining similar or better accuracy.

[LG-131] A Lightweight Federated Learning Approach for Privacy-Preserving Botnet Detection in IoT

链接: https://arxiv.org/abs/2510.03513
作者: Taha M. Mahmoud,Naima Kaabouch
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注: This work has been published in the Proceedings of the 2025 IEEE International Conference on Applied Cloud and Data Science and Applications (ACDSA). The final published version is available via IEEE Xplore at this https URL

点击查看摘要

[LG-132] ask-Level Contrastiveness for Cross-Domain Few-Shot Learning

链接: https://arxiv.org/abs/2510.03509
作者: Kristi Topollai,Anna Choromanska
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-133] D2 Actor Critic: Diffusion Actor Meets Distributional Critic

链接: https://arxiv.org/abs/2510.03508
作者: Lunjun Zhang,Shuo Han,Hanrui Lyu,Bradly C Stadie
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-134] rajectory Data Suffices for Statistically Efficient Policy Evaluation in Finite-Horizon Offline RL with Linear qπ-Realizability and Concentrability

链接: https://arxiv.org/abs/2510.03494
作者: Volodymyr Tkachuk,Csaba Szepesvári,Xiaoqi Tan
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-135] How to Set β_1 β_2 in Adam: An Online Learning Perspective

链接: https://arxiv.org/abs/2510.03478
作者: Quan Nguyen
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注: 15 pages

点击查看摘要

[LG-136] On residual network depth

链接: https://arxiv.org/abs/2510.03470
作者: Benoit Dherin,Michael Munn
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-137] Paris: A Decentralized Trained Open-Weight Diffusion Model

链接: https://arxiv.org/abs/2510.03434
作者: Zhiying Jiang,Raihan Seraj,Marcos Villagra,Bidhan Roy
类目: Graphics (cs.GR); Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-138] LHGEL: Large Heterogeneous Graph Ensemble Learning using Batch View Aggregation ICDM2025

链接: https://arxiv.org/abs/2510.03432
作者: Jiajun Shen,Yufei Jin,Yi He,Xingquan Zhu
类目: Machine Learning (cs.LG)
*备注: Accepted by ICDM 2025

点击查看摘要

[LG-139] Memory-Efficient Backpropagation for Fine-Tuning LLM s on Resource-Constrained Mobile Devices

链接: https://arxiv.org/abs/2510.03425
作者: Congzheng Song,Xinyu Tang
类目: Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-140] raining Variation of Physically-Informed Deep Learning Models

链接: https://arxiv.org/abs/2510.03416
作者: Ashley Lenau,Dennis Dimiduk,Stephen R. Niezgoda
类目: Machine Learning (cs.LG); Materials Science (cond-mat.mtrl-sci)
*备注:

点击查看摘要

[LG-141] Is it Bigger than a Breadbox: Efficient Cardinality Estimation for Real World Workloads

链接: https://arxiv.org/abs/2510.03386
作者: Zixuan Yi,Sami Abu-el-Haija,Yawen Wang,Teja Vemparala,Yannis Chronis,Yu Gan,Michael Burrows,Carsten Binnig,Bryan Perozzi,Ryan Marcus,Fatma Ozcan
类目: Databases (cs.DB); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-142] Estimating link level traffic emissions: enhancing MOVES with open-source data

链接: https://arxiv.org/abs/2510.03362
作者: Lijiao Wang,Muhammad Usama,Haris N. Koutsopoulos,Zhengbing He
类目: Machine Learning (cs.LG); Applications (stat.AP); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-143] High Cycle S-N curve prediction for Al 7075-T6 alloy using Recurrent Neural Networks (RNNs)

链接: https://arxiv.org/abs/2510.03355
作者: Aryan Patel
类目: Machine Learning (cs.LG); Materials Science (cond-mat.mtrl-sci); Applied Physics (physics.app-ph)
*备注:

点击查看摘要

[LG-144] Matching the Optimal Denoiser in Point Cloud Diffusion with (Improved) Rotational Alignment

链接: https://arxiv.org/abs/2510.03335
作者: Ameya Daigavane,YuQing Xie,Bodhi P. Vani,Saeed Saremi,Joseph Kleinhenz,Tess Smidt
类目: Machine Learning (cs.LG); Image and Video Processing (eess.IV)
*备注: under review

点击查看摘要

[LG-145] Semantic-Aware Scheduling for GPU Clusters with Large Language Models

链接: https://arxiv.org/abs/2510.03334
作者: Zerui Wang,Qinghao Hu,Ana Klimovic,Tianwei Zhang,Yonggang Wen,Peng Sun,Dahua Lin
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注:

点击查看摘要

[LG-146] Constant in an Ever-Changing World

链接: https://arxiv.org/abs/2510.03330
作者: Andy Wu,Chun-Cheng Lin,Yuehua Huang,Rung-Tzuo Liaw
类目: Machine Learning (cs.LG)
*备注: in Chinese language

点击查看摘要

[LG-147] Fast frequency reconstruction using Deep Learning for event recognition in ring laser data

链接: https://arxiv.org/abs/2510.03325
作者: Giuseppe Di Somma,Giorgio Carelli,Angela D.V. Di Virgilio,Francesco Fuso,Enrico Maccioni,Paolo Marsili
类目: Machine Learning (cs.LG); Computational Physics (physics.comp-ph); Data Analysis, Statistics and Probability (physics.data-an); Geophysics (physics.geo-ph)
*备注:

点击查看摘要

[LG-148] Attack logics not outputs: Towards efficient robustification of deep neural networks by falsifying concept-based properties

链接: https://arxiv.org/abs/2510.03320
作者: Raik Dankworth,Gesina Schwalbe
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 13 pages, 2 figures, accepted by “7th OVERLAY” workshop

点击查看摘要

[LG-149] SVDefense: Effective Defense against Gradient Inversion Attacks via Singular Value Decomposition

链接: https://arxiv.org/abs/2510.03319
作者: Chenxiang Luo,David K.Y. Yau,Qun Song
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-150] Scaling Laws Revisited: Modeling the Role of Data Quality in Language Model Pretraining

链接: https://arxiv.org/abs/2510.03313
作者: Anirudh Subramanyam,Yuxin Chen,Robert L. Grossman
类目: Machine Learning (cs.LG)
*备注: 18 pages, 6 figures

点击查看摘要

[LG-151] hin Bridges for Drug Text Alignment: Lightweight Contrastive Learning for Target Specific Drug Retrieval

链接: https://arxiv.org/abs/2510.03309
作者: Mallikarjuna Tupakula
类目: Machine Learning (cs.LG); Biomolecules (q-bio.BM)
*备注:

点击查看摘要

[LG-152] Machine Learning Workflows in Climate Modeling: Design Patterns and Insights from Case Studies

链接: https://arxiv.org/abs/2510.03305
作者: Tian Zheng,Subashree Venkatasubramanian,Shuolin Li,Amy Braverman,Xinyi Ke,Zhewen Hou,Peter Jin,Samarth Sanjay Agrawal
类目: Machine Learning (cs.LG); Atmospheric and Oceanic Physics (physics.ao-ph); Applications (stat.AP); Machine Learning (stat.ML)
*备注: Supplement

点击查看摘要

[LG-153] Single-Core Superscalar Optimization of Clifford Neural Layers

链接: https://arxiv.org/abs/2510.03290
作者: X. Angelo Huang,Ruben Ciranni,Giovanni Spadaccini,Carla J. López Zurita
类目: Machine Learning (cs.LG)
*备注: 9 pages

点击查看摘要

[LG-154] Variational Autoencoders-based Detection of Extremes in Plant Productivity in an Earth System Model

链接: https://arxiv.org/abs/2510.03266
作者: Bharat Sharma,Jitendra Kumar
类目: Machine Learning (cs.LG); Methodology (stat.ME); Other Statistics (stat.OT)
*备注:

点击查看摘要

[LG-155] Data-Driven Temperature Modelling of Machine Tools by Neural Networks: A Benchmark

链接: https://arxiv.org/abs/2510.03261
作者: C. Coelho,M. Hohmann,D. Fernández,L. Penter,S. Ihlenfeldt,O. Niggemann
类目: Machine Learning (cs.LG); Computational Engineering, Finance, and Science (cs.CE)
*备注:

点击查看摘要

[LG-156] Adversarial training with restricted data manipulation

链接: https://arxiv.org/abs/2510.03254
作者: David Benfield,Stefano Coniglio,Phan Tu Vuong,Alain Zemkoho
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注: 21 page, 5 figures

点击查看摘要

[LG-157] Light Differentiable Logic Gate Networks

链接: https://arxiv.org/abs/2510.03250
作者: Lukas Rüttgers,Till Aczel,Andreas Plesner,Roger Wattenhofer
类目: Machine Learning (cs.LG); Performance (cs.PF)
*备注:

点击查看摘要

[LG-158] Bayesian Distributional Models of Executive Functioning

链接: https://arxiv.org/abs/2510.00387
作者: Robert Kasumba,Zeyu Lu,Dom CP Marticorena,Mingyang Zhong,Paul Beggs,Anja Pahor,Geetha Ramani,Imani Goffney,Susanne M Jaeggi,Aaron R Seitz,Jacob R Gardner,Dennis L Barbour
类目: Machine Learning (cs.LG); Human-Computer Interaction (cs.HC)
*备注: 42 pages, 8 figures, 1 table

点击查看摘要

[LG-159] A Unified Optimization Framework for Multiclass Classification with Structured Hyperplane Arrangements

链接: https://arxiv.org/abs/2510.05047
作者: Víctor Blanco,Harshit Kothari,James Luedtke
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注: 28 pages, 2 tables, 9 figures

点击查看摘要

[LG-160] Causal Abstractions Categorically Unified

链接: https://arxiv.org/abs/2510.05033
作者: Markus Englberger,Devendra Singh Dhami
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-161] Curiosity-Driven Co-Development of Action and Language in Robots Through Self-Exploration

链接: https://arxiv.org/abs/2510.05013
作者: Theodore Jerome Tinker,Kenji Doya,Jun Tani
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 26 pages, 14 pages of supplementary material

点击查看摘要

[LG-162] Pivotal CLTs for Pseudolikelihood via Conditional Centering in Dependent Random Fields

链接: https://arxiv.org/abs/2510.04972
作者: Nabarun Deb
类目: atistics Theory (math.ST); Machine Learning (cs.LG); Probability (math.PR)
*备注: 73 pages, 1 figure

点击查看摘要

[LG-163] Set to Be Fair: Demographic Parity Constraints for Set-Valued Classification

链接: https://arxiv.org/abs/2510.04926
作者: Eyal Cohen(LPSM (UMR_8001)),Christophe Denis(SAMM),Mohamed Hebiri(LAMA)
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-164] A Noise Resilient Approach for Robust Hurst Exponent Estimation

链接: https://arxiv.org/abs/2510.04811
作者: Malith Premarathna(1),Fabrizio Ruggeri(2),Dixon Vimalajeewa(1) ((1) Department of Statistics, University of Nebraska-Lincoln, (2) CNR IMATI, Milano)
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-165] Kernel ridge regression under power-law data: spectrum and generalization

链接: https://arxiv.org/abs/2510.04780
作者: Arie Wortsman,Bruno Loureiro
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-166] Predictive economics: Rethinking economic methodology with machine learning

链接: https://arxiv.org/abs/2510.04726
作者: Miguel Alves Pereira
类目: General Economics (econ.GN); Machine Learning (cs.LG)
*备注: 8 pages

点击查看摘要

[LG-167] Gini-based Model Monitoring: A General Framework with an Application to Non-life Insurance Pricing

链接: https://arxiv.org/abs/2510.04556
作者: Alexej Brauer,Paul Menzel
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Statistics Theory (math.ST); Statistical Finance (q-fin.ST); Applications (stat.AP)
*备注:

点击查看摘要

[LG-168] Learning Linear Regression with Low-Rank Tasks in-Context

链接: https://arxiv.org/abs/2510.04548
作者: Kaito Takanami,Takashi Takahashi,Yoshiyuki Kabashima
类目: Disordered Systems and Neural Networks (cond-mat.dis-nn); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-169] Quantum generative model on bicycle-sharing system and an application

链接: https://arxiv.org/abs/2510.04512
作者: Fumio Nemoto,Nobuyuki Koike,Daichi Sato,Yuuta Kawaai,Masayuki Ohzeki
类目: Quantum Physics (quant-ph); Machine Learning (cs.LG)
*备注: 8 pages, 11 figures

点击查看摘要

[LG-170] Benchmarking atmospheric circulation variability in an AI emulator ACE2 and a hybrid model NeuralGCM

链接: https://arxiv.org/abs/2510.04466
作者: Ian Baxter,Hamid Pahlavan,Pedram Hassanzadeh,Katharine Rucker,Tiffany Shaw
类目: Atmospheric and Oceanic Physics (physics.ao-ph); Machine Learning (cs.LG)
*备注: 12 pages, 4 main figures, 6 supplementary figures

点击查看摘要

[LG-171] Perspectives on Stochastic Localization

链接: https://arxiv.org/abs/2510.04460
作者: Bobby Shi,Kevin Tian,Matthew S. Zhang
类目: Probability (math.PR); Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG); Statistics Theory (math.ST)
*备注:

点击查看摘要

[LG-172] Zeroth-Order Methods for Stochastic Nonconvex Nonsmooth Composite Optimization

链接: https://arxiv.org/abs/2510.04446
作者: Ziyi Chen,Peiran Yu,Heng Huang
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-173] spd-metrics-id: A Python Package for SPD-Aware Distance Metrics in Connectome Fingerprinting and Beyond

链接: https://arxiv.org/abs/2510.04438
作者: Kaosar Uddin
类目: Computation (stat.CO); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-174] Learning Survival Models with Right-Censored Reporting Delays

链接: https://arxiv.org/abs/2510.04421
作者: Yuta Shikuri,Hironori Fujisawa
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Statistics Theory (math.ST)
*备注: 21 pages, 3 figures, 4 tables

点击查看摘要

[LG-175] Modular and Adaptive Conformal Prediction for Sequential Models via Residual Decomposition

链接: https://arxiv.org/abs/2510.04406
作者: William Zhang,Saurabh Amin,Georgia Perakis
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 11 pages, (37 with appendix), 15 figures

点击查看摘要

[LG-176] CR-EML: Explainable Model Layers for TCR-pMHC Prediction

链接: https://arxiv.org/abs/2510.04377
作者: Jiarui Li,Zixiang Yin,Zhengming Ding,Samuel J. Landry,Ramgopal R. Mettu
类目: Quantitative Methods (q-bio.QM); Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-177] Quantizer Design for Finite Model Approximations Model Learning and Quantized Q-Learning for MDPs with Unbounded Spaces

链接: https://arxiv.org/abs/2510.04355
作者: Osman Bicer,Ali D. Kara,Serdar Yuksel
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-178] Adaptive Coverag e Policies in Conformal Prediction

链接: https://arxiv.org/abs/2510.04318
作者: Etienne Gauthier,Francis Bach,Michael I. Jordan
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: Code at: this https URL

点击查看摘要

[LG-179] Relative Information Gain and Gaussian Process Regression

链接: https://arxiv.org/abs/2510.04277
作者: Hamish Flynn
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 28 pages

点击查看摘要

[LG-180] A Universal Deep Learning Force Field for Molecular Dynamic Simulation and Vibrational Spectra Prediction

链接: https://arxiv.org/abs/2510.04227
作者: Shengjiao Ji,Yujin Zhang,Zihan Zou,Bin Jiang,Jun Jiang,Yi Luo,Wei Hu
类目: Chemical Physics (physics.chem-ph); Machine Learning (cs.LG)
*备注: 19 pages, 5 figures

点击查看摘要

[LG-181] Drax: Speech Recognition with Discrete Flow Matching

链接: https://arxiv.org/abs/2510.04162
作者: Aviv Navon,Aviv Shamsian,Neta Glazer,Yael Segal-Feldman,Gill Hetz,Joseph Keshet,Ethan Fetaya
类目: Audio and Speech Processing (eess.AS); Machine Learning (cs.LG); Sound (cs.SD)
*备注:

点击查看摘要

[LG-182] Simulation-based inference via telescoping ratio estimation for trawl processes

链接: https://arxiv.org/abs/2510.04042
作者: Dan Leonte,Raphaël Huser,Almut E. D. Veraart
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Methodology (stat.ME)
*备注:

点击查看摘要

[LG-183] Self-Speculative Masked Diffusions

链接: https://arxiv.org/abs/2510.03929
作者: Andrew Campbell,Valentin De Bortoli,Jiaxin Shi,Arnaud Doucet
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 32 pages, 7 figures, 3 tables

点击查看摘要

[LG-184] Optimal Computation from Fluctuation Responses

链接: https://arxiv.org/abs/2510.03900
作者: Jinghao Lyu,Kyle J. Ray,James P. Crutchfield
类目: atistical Mechanics (cond-mat.stat-mech); Machine Learning (cs.LG)
*备注: 10 pages, 6 figures; this https URL

点击查看摘要

[LG-185] Spectral Thresholds for Identifiability and Stability:Finite-Sample Phase Transitions in High-Dimensional Learning

链接: https://arxiv.org/abs/2510.03809
作者: William Hao-Cheng Huang
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:

点击查看摘要

Abstract:In high-dimensional learning, models remain stable until they collapse abruptly once the sample size falls below a critical level. This instability is not algorithm-specific but a geometric mechanism: when the weakest Fisher eigendirection falls beneath sample-level fluctuations, identifiability fails. Our Fisher Threshold Theorem formalizes this by proving that stability requires the minimal Fisher eigenvalue to exceed an explicit O(\sqrtd/n) bound. Unlike prior asymptotic or model-specific criteria, this threshold is finite-sample and necessary, marking a sharp phase transition between reliable concentration and inevitable failure. To make the principle constructive, we introduce the Fisher floor, a verifiable spectral regularization robust to smoothing and preconditioning. Synthetic experiments on Gaussian mixtures and logistic models confirm the predicted transition, consistent with d/n scaling. Statistically, the threshold sharpens classical eigenvalue conditions into a non-asymptotic law; learning-theoretically, it defines a spectral sample-complexity frontier, bridging theory with diagnostics for robust high-dimensional inference.

[LG-186] A Benchmark Study of Deep Learning Methods for Multi-Label Pediatric Electrocardiogram-Based Cardiovascular Disease Classification

链接: https://arxiv.org/abs/2510.03780
作者: Yiqiao Chen
类目: ignal Processing (eess.SP); Machine Learning (cs.LG)
*备注: 8 pages, 5 figures

点击查看摘要

[LG-187] he analogy theorem in Hoare logic

链接: https://arxiv.org/abs/2510.03685
作者: Nikitin Nikita
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Logic (math.LO); Computation (stat.CO); Methodology (stat.ME)
*备注:

点击查看摘要

[LG-188] Composite Optimization with Error Feedback: the Dual Averag ing Approach

链接: https://arxiv.org/abs/2510.03507
作者: Yuan Gao,Anton Rodomanov,Jeremy Rack,Sebastian Stich
类目: Optimization and Control (math.OC); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-189] Quantum feature-map learning with reduced resource overhead

链接: https://arxiv.org/abs/2510.03389
作者: Jonas Jäger,Philipp Elsässer,Elham Torabian
类目: Quantum Physics (quant-ph); Machine Learning (cs.LG)
*备注: 17 pages, 9 figures

点击查看摘要

[LG-190] Bias and Coverag e Properties of the WENDy-IRLS Algorithm

链接: https://arxiv.org/abs/2510.03365
作者: Abhi Chawla,David M. Bortz,Vanja Dukic
类目: Methodology (stat.ME); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:

点击查看摘要

[LG-191] Assessing the impact of contact time on leachate chemistry from recycled concrete aggregates

链接: https://arxiv.org/abs/2510.03344
作者: Morgan D. Sanger,Gabrielle Campagnola,Robin Ritchey,Tuncer B. Edil,Matthew Ginder-Vogel
类目: Chemical Physics (physics.chem-ph); Materials Science (cond-mat.mtrl-sci); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-192] Machine Learning and Control: Foundations Advances and Perspectives

链接: https://arxiv.org/abs/2510.03303
作者: Enrique Zuazua
类目: Optimization and Control (math.OC); Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注:

点击查看摘要

[LG-193] Mathematically rigorous proofs for Shapley explanations

链接: https://arxiv.org/abs/2510.03281
作者: David van Batenburg
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:

点击查看摘要

[LG-194] Quantile-Scaled Bayesian Optimization Using Rank-Only Feedback

链接: https://arxiv.org/abs/2510.03277
作者: Tunde Fahd Egunjobi
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Statistics Theory (math.ST)
*备注: 28 pages, 7 figures

点击查看摘要

[LG-195] Improving SP 500 Volatility Forecasting through Regime-Switching Methods

链接: https://arxiv.org/abs/2510.03236
作者: Ava C. Blake,Nivika A. Gandhi,Anurag R. Jakkula
类目: atistical Finance (q-fin.ST); Machine Learning (cs.LG); Econometrics (econ.EM)
*备注:

点击查看摘要

信息检索

[IR-0] opic-Specific Classifiers are Better Relevance Judges than Prompted LLM s

链接: https://arxiv.org/abs/2510.04633
作者: Lukas Gienapp,Martin Potthast,Harrisen Scells,Eugene Yang
类目: Information Retrieval (cs.IR)
*备注: 15 pages, 3 figures, 2 tables

点击查看摘要

[IR-1] MARCO: A Cooperative Knowledge Transfer Framework for Personalized Cross-domain Recommendations SIGIR

链接: https://arxiv.org/abs/2510.04508
作者: Lili Xie,Yi Zhang,Ruihong Qiu,Jiajun Liu,Sen Wang
类目: Information Retrieval (cs.IR)
*备注: SIGIR-AP 2025

点击查看摘要

[IR-2] Evaluating Keyframe Layouts for Visual Known-Item Search in Homogeneous Collections

链接: https://arxiv.org/abs/2510.04396
作者: Bastian Jäckl,Jiří Kruchina,Lucas Joos,Daniel A. Keim,Ladislav Peška,Jakub Lokoč
类目: Multimedia (cs.MM); Information Retrieval (cs.IR)
*备注: 28 Pages, 17 Figures

点击查看摘要

[IR-3] he LCLStream Ecosystem for Multi-Institutional Dataset Exploration

链接: https://arxiv.org/abs/2510.04012
作者: David Rogers,Valerio Mariani,Cong Wang,Ryan Coffee,Wilko Kroeger,Murali Shankar,Hans Thorsten Schwander,Tom Beck,Frédéric Poitevin,Jana Thayer
类目: Information Retrieval (cs.IR); Instrumentation and Detectors (physics.ins-det)
*备注: 3 figures

点击查看摘要

[IR-4] Evaluating High-Resolution Piano Sustain Pedal Depth Estimation with Musically Informed Metrics

链接: https://arxiv.org/abs/2510.03750
作者: Hanwen Zhang,Kun Fang,Ziyu Wang,Ichiro Fujinaga
类目: Information Retrieval (cs.IR); Sound (cs.SD); Audio and Speech Processing (eess.AS)
*备注:

点击查看摘要

附件下载

点击下载今日全部论文列表

Arxiv今日论文 | 2025-10-07

目录

概览 (2025-10-07)

自然语言处理

计算机视觉

人工智能

机器学习

信息检索

附件下载